• Google’s AMIE system matched or beat primary care physicians on clinical management reasoning across 100 multi-visit scenarios. Published in Nature, and the study design is more rigorous than most in this space.

• CANVAS turns standard H&E slides into virtual spatial proteomics maps, predicting tumour microenvironment neighbourhoods from cheap histology. Validated across 5,000 patients and 9 cancer types.

• AMPGAN v3 is the first generative model for antimicrobial peptides that handles non-canonical amino acids and chemical modifications. Two of five generated candidates showed real antimicrobial activity. They also built an agentic pipeline around it, which is where this gets interesting.

Kiin Pioneer Programme

We built a platform that helps researchers speed up their entire science, from literature review and biomarker discovery to bioinformatics and computational chemistry. If your workflow involves pulling findings from five different places before you can actually act on any of them, this is for that.

The Pioneer Programme gives academic labs and non-profits one year of free access, plus support from our science team. No cost, no data transfer, all IP stays with your institution. Applications close August, cohort starts September.

Read more about the programme

Apply now

AMIE: Conversational AI for disease management

🧪 Where This Fits

Most AI-in-medicine papers test whether a model can diagnose from a static vignette. That is a solved problem at this point, or at least a well-explored one. What has not been tested seriously is whether an AI system can manage a patient over time: adjust treatment plans across multiple visits, respond to new lab results, and prescribe medications safely. That is what primary care actually involves, and it is where AMIE (Articulate Medical Intelligence Explorer) from Google DeepMind now enters.

Previous AMIE work showed the system could match physicians on diagnostic conversations. This paper extends it to management reasoning, which is harder because it involves sequential decisions, guideline interpretation, and medication safety. The comparison set is interesting: 21 board-certified primary care physicians across 100 multi-visit case scenarios grounded in UK NICE and BMJ Best Practice guidelines.

🔍 What It Is

• Management reasoning for longitudinal care is largely untested for AI. Liévin et al. from Google DeepMind present an agentic version of AMIE optimised for multi-visit clinical management and medication reasoning.

• Uses Gemini’s long-context window to combine in-context retrieval of clinical guidelines with structured reasoning. A multi-agent architecture handles dialogue, management planning, and medication lookup from drug formularies (OpenFDA, BNF).

• In a blinded virtual OSCE study, AMIE scored significantly higher than PCPs on treatment preciseness (96% vs. 62%, p<0.001) and guideline alignment (93% vs. 75% by visit 3). On the RxQA medication benchmark, AMIE outperformed PCPs on harder questions (57.9% vs. 47.8%, p<0.001). Specialist physicians and patient actors preferred AMIE 47% of the time vs. 7% for PCPs.

💡 Why This Is Cool

The honest reaction here is: this is impressive and somewhat uncomfortable. AMIE is not just pattern-matching against guidelines, it is reasoning about how to adjust a plan given what happened at the last visit. The study design (blinded OSCE, specialist evaluators, real guidelines) is more credible than most in this space. The medication reasoning results are particularly notable because prescribing errors are a leading cause of preventable harm. That said, this is still a virtual scenario, not a real clinic with real patients who do unexpected things. The gap between “performs well in structured evaluation” and “can safely manage my mum’s hypertension” remains large. What this does establish is that the technical capability exists. The regulatory and deployment questions are now the binding constraint, not the model performance.

📃 Read the paper.

💻 Try the code.

CANVAS: Virtual spatial tumor profiling from histopathology

🧪 Where This Fits

Spatial proteomics (CODEX, MIBI, etc.) can map the tumour microenvironment at single-cell resolution, but it costs thousands per sample and requires specialised equipment. Standard H&E histopathology costs almost nothing and is already collected for every cancer patient. The question is whether you can infer the spatial biology from the cheap stain. Previous attempts have tried to predict individual protein expression from H&E, but that approach is fragile: sensitive to staining variation, limited to a handful of markers, and accuracy drops quickly. CANVAS takes a different approach, predicting cellular neighbourhood patterns rather than individual proteins, which is a more robust prediction target.

This connects to a broader trend in computational pathology where foundation models (UNI, Virchow, CONCH) have made feature extraction from H&E much more powerful, and the question is now what downstream tasks those features can support. CANVAS uses these pretrained features to bridge modalities rather than training from scratch.

🔍 What It Is

• Spatial proteomics reveals the tumour microenvironment in detail but costs too much to scale. Li et al. from Stanford present CANVAS, an AI platform that predicts spatial cellular neighbourhood structures directly from standard H&E slides, trained on an atlas of 18 million cells profiled by 41-plex CODEX imaging.

• CANVAS defines 10 reproducible cellular neighbourhoods from CODEX data across 457 lung cancer patients, then trains a pathology foundation model to predict these neighbourhood patterns from co-registered H&E images. It operates at the ecological niche level rather than individual cell types.

• Applied to over 5,000 patients across 9 cancer types, CANVAS-derived spatial features predicted immunotherapy response with AUCs above 0.75 at 6, 12, and 24 months. The spatial signature stratified patients by progression-free survival (HR = 2.42, p<0.001) and outperformed established biomarkers including TMB, PD-L1 expression, and TLS. Validated externally on a Cancer Moonshot Biobank cohort.

💡 Why This Is Cool

This matters for a specific reason: immunotherapy response prediction is a clinical problem where existing biomarkers (PD-L1, TMB) work poorly. About 20-30% of patients respond to checkpoint inhibitors, and we are bad at predicting who they will be beforehand. CANVAS proposes that the spatial organisation of the tumour microenvironment, inferred from a slide that already exists in every pathology lab, is more informative than the molecular markers we have been relying on. If the external validation holds up across broader cohorts (the Moonshot cohort is small at n=40), this could actually change who gets prescribed immunotherapy. The non-commercial license limits immediate industry adoption, but for academic cancer centres this is usable now.

📃 Read the paper.

💻 Try the code.

AMPGAN v3: Agentic discovery of non-canonical antimicrobial peptides

🧪 Where This Fits

Antimicrobial resistance causes over a million deaths annually, and no new antibiotic class has been commercialised since 2000. Antimicrobial peptides (AMPs) are attractive because they disrupt membranes through physical interactions, making resistance harder to develop. Generative models for AMP design exist (PepGAN, HydrAMP, AMP-Designer), but they all share two limitations: they only work with natural L-amino acids, and they require manual filtering of outputs. Real therapeutic peptides need D-amino acids and terminal modifications to survive in the body. AMPGAN v3 is the first generative model that handles these non-canonical chemistries, and PepCraft wraps it in a multi-agent pipeline that automates the filtering.

The field has been generating lots of candidate peptides computationally, but the translation gap to actual antimicrobials has been wide. Most papers stop at predicted activity scores. This one synthesises candidates and tests them against real bacteria, which is the bar that matters.

🔍 What It Is

• Generative models for antimicrobial peptides are limited to natural amino acids and produce outputs that need extensive manual curation. Jung et al. from the University of Vermont and Purdue present AMPGAN v3, a conditional GAN that generates antimicrobial peptides with D-amino acids and terminal modifications, paired with PepCraft, a multi-agent framework for automated AMP discovery.

• AMPGAN v3 separates adversarial training across two discriminators: one for sequence realism, one for antimicrobial activity prediction. This fixes the training instability that plagued earlier versions (only ~10% of AMPGAN v2 runs produced usable models). PepCraft uses a Planning Agent to coordinate specialised executors for generation, physicochemical filtering, and database verification.

• Two of five synthesised candidates showed clear antimicrobial activity against Gram-positive strains, with the best reaching MIC of 8 μg/mL against B. subtilis. The candidates spanned three structural classes (alpha-helical, beta-hairpin, random coil) and incorporated D-amino acids and amidation, which previous generative methods cannot produce. PepCraft’s prioritisation recommendations aligned with the wet-lab results.

💡 Why This Is Cool

The wet-lab hit rate (2/5) is respectable for a generative model, and the chemical space expansion is the real contribution. Every other AMP generator is restricted to natural amino acids, which means their outputs degrade rapidly in serum. Expanding the vocabulary to include D-amino acids and terminal caps makes the generated peptides actually viable as therapeutics rather than just interesting sequences. The agentic pipeline (PepCraft) is early-stage and exploratory, but it points toward a pattern we will see more of: generative models wrapped in verification agents that can filter, cross-reference, and prioritise without human intervention. This was accepted at the ICML 2026 GenBio workshop, not a top venue, and the validation is limited to Gram-positive bacteria. Worth watching, not yet proven at scale.

📃 Read the paper.

💻 Try the code.

🗓️ Events & Competitions

The best competitions, hackathons, and community challenges in AI x life sciences, curated weekly. Know something worth featuring? Reply and let us know.

More upcoming events:

Creative Disruption Forum: Modern Drug Discovery | June 18, NIAB Cambridge

A full-day forum for biotech and R&D leaders exploring how technology is changing small molecule drug discovery. Keynote interviews with industry thought leaders followed by workshops under Chatham House Rules, limited to 60 attendees. Part of Cambridge Wide Open Week. Organised by Graham Combe and Prof Tony Sedgwick. £60 for biotech companies.

London Protein Design Day | June 23, Imperial College London

The first edition of a one-day symposium bringing together London’s protein design community and beyond. Programme spans AI-driven design, molecular dynamics, and bioinformatics, with applications across enzymes, antibodies, and materials. Organised by Pietro Sormanni, Rebecca Birolo, and Jakub Lála. In person only.

BioHackathon Europe 2026 | November 9-13, Barcelona

ELIXIR’s annual international bioinformatics hackathon, running since 2018. 160+ participants, five days of collaborative coding on open bioinformatics infrastructure and tools. The call for project proposals has now closed.

Thanks for reading!

💬 Get involved

We’re always looking to grow our community. If you’d like to get involved, contribute ideas or share something you’re building, fill out this form or reach out to me directly.

Connect With Us

Have questions or suggestions? We'd love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

• mRNAutilus generates entire therapeutic mRNA sequences from scratch, and the wet-lab numbers are hard to argue with: 400x over wild-type expression, beating commercial Spike constructs.

• EpiFormer brings geometric deep learning to epitope prediction with a 40% F1 boost. A nice complement to last week’s ESM binder design coverage, now from the antigen side.

• Folddisco indexes 53 million protein structures and finds structural motifs in seconds. The Steinegger lab keeps quietly building infrastructure that makes everyone else’s work faster.

Kiin Pioneer Programme

We built a platform that helps researchers speed up their entire science, from literature review and biomarker discovery to bioinformatics and computational chemistry. If your workflow involves pulling findings from five different places before you can actually act on any of them, this is for that.

The Pioneer Programme gives academic labs and non-profits one year of free access, plus support from our science team. No cost, no data transfer, all IP stays with your institution. Applications close August, cohort starts September.

Read more about the programme

Apply now

mRNAutilus: Multi-Objective-Guided Discrete Generation of mRNA with Optimized Therapeutic Properties

🔬 Designing full-length therapeutic mRNAs means optimising stability, translation efficiency, and codon usage simultaneously. Current methods tackle these objectives piecemeal, stitching together separately optimised UTRs and coding regions, which leaves performance on the table.

Patel et al. from the Chatterjee lab at Duke present mRNAutilus, a generative framework that designs complete mRNA transcripts optimised across multiple properties at once.

🧬 The system trains a masked discrete diffusion model on millions of full-length mRNAs, then steers generation with Monte Carlo tree guidance to hit multiple objectives without retraining. It operates on whole transcripts rather than modular components.

⚡ Zero-shot mRNAutilus designs encoding firefly luciferase achieved over 400-fold higher expression than wild-type, outperforming commercial baselines. For SARS-CoV-2 Spike, designs matched or surpassed both clinically used constructs and lab-optimised sequences. The framework also generalised to prime editing guides and targeted protein degradation, which suggests this is not a one-trick benchmark result.

🧪 Where This Fits

mRNA therapeutics have had a sequencing problem disguised as a design problem. Post-COVID, the bottleneck is no longer “can we make mRNA drugs?” but “can we make them well enough to compete on potency and manufacturing cost?” Tools like LinearDesign (Zhang lab, 2023) optimise coding sequences for stability, and UTR-focused approaches pick regulatory elements, but these treat the transcript as a set of independent modules. mRNAutilus is the first framework I have seen that treats the entire transcript as a single generative object, which matters because interactions between UTRs and coding regions affect folding and translation in ways modular approaches miss.

The wet-lab validation is what separates this from yet another generative model paper. Beating commercial constructs for Spike expression is a meaningful bar, not an in silico benchmark. The generalisation to prime editing and degraders suggests the architecture is flexible enough to not be overfit to reporter assays. The timing makes sense too: masked diffusion models have matured enough (thanks to protein and genomics applications) that applying them to mRNA sequences is a natural next step, and the Monte Carlo tree guidance borrows from AlphaGo-era decision strategies to handle multi-objective trade-offs without expensive retraining.

For readers working in mRNA therapeutics: this is worth watching closely. The code is not yet public, which limits immediate adoption, but the approach could compress the design-test cycle considerably once available.

💡 Why This Is Cool

The shift from “optimise one property” to “generate the whole thing optimised” matters more than it sounds. The history of biologics design is littered with tools that optimised one metric while inadvertently breaking another. If multi-objective generation holds up across more constructs and delivery contexts, it moves mRNA design closer to what protein design achieved with diffusion models over the past two years. The open question is whether the approach scales to longer, more complex transcripts and novel target classes beyond the well-studied ones shown here.

📃 Read the paper.

EpiFormer: Learning Antigen-Antibody Interactions for Epitope Prediction via Geometric Deep Learning

🔬 Predicting which surface residues an antibody will target on an antigen remains stubbornly difficult. Most existing methods treat the antigen in isolation, ignoring the antibody entirely or bolting antibody information on as a late-stage afterthought.

Ahmed et al. from Georgia State University introduce EpiFormer, a geometric deep learning framework that models antigen-antibody interactions through interleaved cross-attention within GNN encoding layers.

🧬 Rather than encoding antigen and antibody separately then combining representations at the end, EpiFormer threads cross-attention between the two structures at every encoding layer. This allows bidirectional information flow throughout the representation, so the model learns how antibody geometry constrains which epitope residues are accessible.

⚡ On standard benchmarks, EpiFormer achieves over 40% improvement in F1 score compared to previous best methods. That is a substantial jump for a prediction task where incremental gains of 2-5% have been typical. The model operates on 3D structural inputs from antibody-antigen complexes.

🧪 Where This Fits

Epitope prediction sits upstream of antibody engineering: if you know where an antibody binds, you can design better binders and prioritise vaccine targets. Previous approaches like DiscoTope and ElliPro use geometry and surface properties of the antigen alone, which is a bit like predicting where a key fits without looking at the lock. More recent methods (PECAN, AsEP) incorporated paratope information but typically as a separate encoding step with late fusion.

EpiFormer’s contribution is architectural rather than data-driven. The interleaved cross-attention ensures antibody context informs antigen representations from the start rather than being concatenated at the decision layer. This connects naturally to last week’s coverage of ESM-based binder design: that work generates antibodies given a target, while EpiFormer predicts where on the target those antibodies will land. Together they cover both directions of the same binding prediction problem.

The 40% F1 improvement is striking, though it warrants some caution. Epitope prediction benchmarks are notoriously sensitive to train/test splitting, and structural epitope datasets remain small (a few thousand complexes in SAbDab). Whether this holds on truly novel antigen folds or just reflects better exploitation of known structural patterns is an open question. The code is available, which helps.

💡 Why This Is Cool

The “interleave information early rather than fuse late” lesson keeps appearing across structural biology. AlphaFold did it for MSA and structure tracks. ESM3 does it for sequence, structure, and function. EpiFormer applies the same intuition to a paired prediction problem. The field has been underestimating how much cross-modal information gets lost in late-fusion architectures, and each new result in this direction makes that clearer. For antibody discovery teams, this is immediately useful if it generalises beyond the benchmark setting.

📃 Read the paper.

💻 Try the code.

Structural Motif Search Across the Protein Universe with Folddisco

🔬 Finding recurring 3D structural motifs (zinc fingers, catalytic triads, protein-protein interaction surfaces) across millions of predicted structures is computationally prohibitive. Existing methods either cannot handle discontinuous motifs or choke on databases beyond a few hundred thousand structures.

Kim et al. from the Steinegger lab at Seoul National University present Folddisco, a structural motif search tool that indexes 53 million AFDB50 structures in a 1.45 TB index and returns query results in seconds.

🧬 Folddisco encodes proximal residue pairs into geometric feature sets (distances, angles, and side-chain orientation via torsion angles), stores them in a position-independent inverted index, and ranks hits using an IDF-based coverage score that rewards rare features. This handles both short continuous motifs and long discontinuous ones.

⚡ Indexing is 11x faster to build and 4x more storage-efficient than previous state-of-the-art. Query speed is 20-fold faster than pyScoMotif on the full pipeline. On the zinc finger benchmark against the human proteome, Folddisco outperformed both RCSB and pyScoMotif on recall while maintaining higher precision. It also successfully distinguished active from inactive GPCR conformational states using activation motifs.

🧪 Where This Fits

This is infrastructure work, and the most consequential kind. The AlphaFold database gave us 200+ million predicted structures, but searching them structurally has lagged far behind searching them by sequence (where tools like Foldseek, also from the Steinegger lab, already operate at scale). Folddisco fills the motif-search gap: given a 3D pattern of interest, find everywhere it occurs across all known and predicted protein structures.

The practical value becomes clear with the GPCR example. Being able to query “show me all structures with this activation motif” across both experimental PDB structures and AlphaFold predictions means you can study conformational states at proteome scale. That was previously manual curation work. Similarly, the zinc finger detection in uncharacterised metagenomic proteins (from ESM30) demonstrates functional annotation where sequence-based methods fail entirely.

Folddisco’s main limitation is its 20-angstrom connectivity constraint, which means it cannot detect spatially distant functional sites like remote allosteric pockets. The IDF scoring also struggles with very short motifs. These are known trade-offs for the speed gains.

💡 Why This Is Cool

The Steinegger lab has been building the search infrastructure for the structure-prediction era piece by piece: MMseqs2 for sequences, Foldseek for structure alignment, and now Folddisco for motif search. Each tool makes the previous one more useful. What matters here is not the individual benchmarks but the fact that motif search at 53-million-structure scale is now a webserver query rather than a compute cluster job. That means anyone with a structural intuition and a browser can generate hypotheses that previously required a compute cluster and custom code. The webserver is live at search.foldseek.com/folddisco.

📃 Read the paper. 💻 Try the code.

🗓️ Events & Competitions

The best competitions, hackathons, and community challenges in AI x life sciences, curated weekly. Know something worth featuring? Reply and let us know.

More upcoming events:

Creative Disruption Forum: Modern Drug Discovery | June 18, NIAB Cambridge

A full-day forum for biotech and R&D leaders exploring how technology is changing small molecule drug discovery. Keynote interviews with industry thought leaders followed by workshops under Chatham House Rules, limited to 60 attendees. Part of Cambridge Wide Open Week. Organised by Graham Combe and Prof Tony Sedgwick. £60 for biotech companies.

London Protein Design Day | June 23, Imperial College London

The first edition of a one-day symposium bringing together London’s protein design community and beyond. Programme spans AI-driven design, molecular dynamics, and bioinformatics, with applications across enzymes, antibodies, and materials. Organised by Pietro Sormanni, Rebecca Birolo, and Jakub Lála. Abstract deadline for poster/oral presentations is this Saturday (May 17). In person only.

BioHackathon Europe 2026 | November 9-13, Barcelona

ELIXIR’s annual international bioinformatics hackathon, running since 2018. 160+ participants, five days of collaborative coding on open bioinformatics infrastructure and tools. The call for project proposals opens March 16 and closes April 15 - so if you want to lead a project, that’s your window.

Thanks for reading!

💬 Get involved

We’re always looking to grow our community. If you’d like to get involved, contribute ideas or share something you’re building, fill out this form or reach out to me directly.

Connect With Us

Have questions or suggestions? We'd love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

Who this piece is for: Computational biology professionals and hiring managers navigating the UK market in 2026.

What this covers: Where the roles are, what they pay, and what actually gets people hired, based on recruiter data and 750+ live listings.

The takeaway: The market rewards specialists who can ship, not generalists who can apply. Infrastructure roles are where demand is highest, entry-level is brutally oversaturated, and your visibility matters more than your credentials.

Freebie alert: We know how hard science is. That’s why we built the Pioneer Programme.

We’re selecting academic and nonprofit teams to get one year of free access to our drug discovery platform, with support from our science team. If you spend more time pulling together findings from different sources than actually acting on them, it’s worth applying.

No cost, no data transfer, all IP stays with your institution. Applications close August, cohort starts September.

Read more about the programme

Apply now

The computational biology job market in 2026 has split in two. We wanted to dive deeper to understand why, and try to think about what might happen next.

Traditional pharma has been making considerable cuts for over a year. Bayer alone cut roughly 14,500 roles between 2022 and 2026, with the steepest reductions in 2024 and 2025. BMS and Teva followed with thousands more. AI-native biotech, meanwhile, is hiring faster than the talent pool can seem to keep up. As a quick example, Isomorphic Labs has 21 open ML drug discovery roles in London alone.

We looked into the market dynamics, skill levels, salary benchmarks, and hiring patterns across 2,000+ UK listings on LinkedIn, and spoke to Joe Phillips, Principal Consultant BioAI at Cubiq Recruitment, a specialist recruiter in bio-AI and computational life sciences. Everything we found is in here: where the roles are, what they pay, who’s getting hired, and what we think happens next.

“There’s an enormous amount of strong academic talent coming through, but the number of genuinely junior opportunities is tiny compared to demand. The candidates who stand out usually have something beyond their degree alone to show.”

📊 Where the roles are

Across the UK, there are currently 2,000+ open positions in computational biology, bioinformatics, AI drug discovery, and adjacent infrastructure roles (source: LinkedIn Jobs, May 2026). These span big pharma, AI-native biotech startups, NHS trusts, CROs, and platform companies selling into life sciences.

• MLOps and platform engineering (1,000+ roles): The biggest category by far, and probably the most surprising if you haven’t been watching the infrastructure side of biotech. Why so many? McKinsey’s 2025 State of AI report found that 88% of companies now use AI in at least one business function (up from 78% the year before), but roughly two-thirds are still stuck in pilot mode. Companies built research teams over the last few years, proved that their models work, and now need people who can operationalise them at scale. Joe says this is the biggest shift he’s seen: “A lot of this work was previously absorbed by ML Engineers. Now the cost and complexity around compute, GPU utilisation, and inference has become significant enough that firms are hiring specialists.”

• Bioinformatics (355 roles): The broadest category, spanning clinical bioinformatics, genomics, spatial and single-cell analysis. Also the most accessible at entry level: 33% of bioinformatics listings are entry-level, compared to just 12% in ML drug discovery. For early career candidates, this is where the door is most open, across both big pharma and smaller biotech. Startups tend to offer faster progression and broader scope; pharma offers stability and established infrastructure.

• Computational biology (164 roles): Core comp bio and adjacent research scientist roles. Spans both pharma and biotech.

• ML in drug discovery (137 roles): Protein design, ADMET prediction, molecular modelling. Isomorphic Labs dominates with 21 positions, followed by Relation Therapeutics with 16. Newer players like CuspAI and Boltz have also been hiring heavily over the past 12 months. Small in absolute numbers, but likely one of the fastest-growing categories year on year: the AI in drug discovery market is expanding at around 25% annually (Verified Market Research). Even if the number of roles today looks modest, the trajectory and the funding flowing in suggest this will look very different by 2028. The roles that exist here tend to be senior, well compensated, and competitive.

• Clinical AI (~500 roles): The broadest umbrella, covering health informatics through to clinical data science.

• AI protein design (9 roles): Tiny, highly specialised, and almost exclusively mid-senior level.

The other trend Joe flags: forward-deployed engineers and technical client facing hires. “A lot of companies are commercialising scientific ML platforms now rather than running their own therapeutics pipelines, so they need engineers and scientists who can comfortably operate across product, research, and client conversations.” GTM hiring is picking up too as firms move beyond pure research mode.

📍 Geography: the Golden Triangle and beyond

If you’ve been paying attention to the London tech scene, the top of this list won’t surprise you.

“Kings Cross is obviously on fire at the moment,” says Joe. “Oxford and Cambridge are also hotspots, although of the two, Cambridge is always an easier sell to candidates because of the commutability from London.”

Outside the Golden Triangle, the Northern Arc (Leeds, Liverpool, Manchester, Sheffield) is showing up as a secondary cluster, backed by Northern Gritstone funding for life science and deep tech spinouts. Glasgow also appears consistently in bioinformatics listings, driven by NHS Scotland roles. For candidates willing to look beyond the south-east, the cost of living advantage is real, particularly when London salaries don’t always come with a proportional premium.

💰 What it pays

So where do these 2,000+ roles sit on salary? UK data for comp bio is notoriously thin, which makes it hard for candidates to know when an offer is fair. Joe shared benchmarks from his recruitment work:

The ML premium is hard to ignore. A senior ML engineer in biotech can earn nearly double what a senior bioinformatician earns. That comes down to scarcity of strong ML talent and the direct commercial impact these roles carry: GPU optimisation and inference efficiency directly affect a company’s burn rate.

Joe’s caveat: “This can massively depend on the size of the company and what they are working on. There are always outliers.” Community data backs this up. Biotech equity is less liquid than FAANG RSUs, so even when base salaries match, total compensation often lags tech. The trade-off is mission, ownership, and the fact that senior bio-AI roles are closing the gap faster than any other life sciences category.

🏠 Remote, hybrid, or on-site?

Many comp bio professionals came up during the pandemic era of fully remote work, so this question comes up constantly. The answer depends on what you actually do day to day.

Bioinformatics is the outlier for remote work (46%), probably because much of the work is pipeline-based and can run independently. Drug discovery and comp bio skew heavily in-person, particularly at early-stage companies where wet lab and dry lab collaboration matters daily.

“Fully remote roles are still fairly rare,” says Joe. “If anything, more companies are trying to get people together in person more, especially at earlier stages where collaboration between science and engineering matters a lot.”

In practice, remote flexibility correlates strongly with sub-field. Bioinformatics and genomics data science offer the most options; ML drug discovery is almost entirely on-site.

🛠️ Skills that matter now

The roles exist and the salaries are there, the question is really just what actually gets you through the door as the technical bar has moved. Python is the most used programming language in the world (GitHub’s 2024 Octoverse report placed it at number one for the first time, overtaking JavaScript). R remains critical for statistical genomics. Beyond the basics, what actually separates candidates falls into three tiers.

Table stakes (expected, not differentiating):

• Python, R, Bash/Unix

• Git, Docker, AWS or GCP

• PyTorch, scikit-learn

• Nextflow or Snakemake

• Basic ML (regression, classification, clustering)

Differentiators (what gets you to the top of the pile):

• GPU optimisation and distributed systems (the single most in-demand infrastructure skill right now)

• Protein language models (e.g. ESM-2, ProtTrans)

• JAX proficiency (driven by the DeepMind/Isomorphic ecosystem)

• Production ML deployment (Kubernetes, inference optimisation)

• Cross-functional communication: being able to sit across product, research, and client conversations

Emerging (bet on these for 2027):

• Diffusion models for molecular generation (e.g. DiffDock, FrameFlow)

• Geometric deep learning and equivariant neural networks

• LLMs for biomedical data (RAG architectures, agentic AI for research)

• Foundation models for genomics (single-cell, spatial transcriptomics)

PyTorch has decisively won over TensorFlow in bio-AI research. JAX has gone from niche to essential for structural biology. Perl has disappeared from modern curricula entirely. The field moves fast enough that what was cutting-edge in 2023 (basic AlphaFold usage) is now baseline knowledge.

What we think: The agentic AI category is the one to watch. Right now, “agentic” is mostly a buzzword on job listings. Within 18 months it will be a real job requirement, because the companies that figure out how to automate their literature review, hypothesis generation, and experimental design pipelines will move significantly faster than those relying on manual researcher effort. If you’re picking a side project to build in public, an agentic research workflow is probably the highest-signal thing you could show a hiring manager right now.

🎯 How to stand out as a company and an employee (from someone who sees 1,000 CVs)

Knowing the right skills is one thing, betting noticed in a pile of 300 applicants is another. The entry-level bottleneck is worth spelling out from both sides. Candidates are applying into a tiny number of junior roles: only 11-12% of ML drug discovery and comp bio positions are entry-level, yet the pipeline of qualified graduates is enormous. Joe puts numbers to it: a senior role might attract 50 applicants, of which maybe one is genuinely relevant. A junior role pulls 300-350, of which 7-10 are a real fit. The ratio of applicants to roles is 6-7x higher at entry level, and even then, most applications miss the mark. Companies, meanwhile, are drowning in applications and still can’t find the right people. The volume of inbound is high; the signal-to-noise ratio is low.

Joe’s advice on what separates the top candidates:

1. Show evidence of building, not just researching. “Founders and hiring teams often love seeing evidence that someone can actually build and ship things, not just research them theoretically. Particularly in bio, there’s a real appreciation for candidates who can move between experimentation and execution.”

2. Be visible where recruiters actually look. This goes beyond LinkedIn. “A huge chunk of our time is tracking publications and collaborations to map who’s working on what, which labs are producing interesting work. There’s also GitHub and Hugging Face, where we’re digging through models, software tooling and tech stacks.” Active repositories, conference contributions, hackathon results, and community involvement all increase visibility.

3. Go deep in interviews, not broad. “A lot of candidates stay very high-level when explaining their work, but hiring teams usually want to go much deeper than people expect. They want candidates to talk through how decisions were made, why certain approaches worked.” Approach interviews like an external consultant: figure out exactly what you are there to improve.

4. Bridge the academia-industry gap deliberately. “Where people sometimes struggle is around product and commercial awareness. Industry teams aren’t just thinking about whether something works academically. They’re thinking about usability, timelines, scalability, deployment and business impact.” The candidates who transition best have already sought out internships, collaborations, or commercial projects while in academia.

5. Network into the hidden job market. “The people most in-demand tend to have clear signals that they’re genuinely interested in their area outside their day job. They’re naturally around others in the space a lot of the time, so are closer to that hidden job market that’s often rife with word-of-mouth opportunities.”

What we think: The pipeline problem in comp bio is structural, not cyclical. Universities produce graduates faster than the industry can absorb at junior level, while senior roles go unfilled for months. This won’t close by itself. The people who break through are the ones who’ve already demonstrated they can operate above their experience level. A polished GitHub with one well documented, production quality project is worth more than five papers in middling journals. Hiring managers are pattern-matching for “can this person ship something on day one,” and the evidence needs to be visible before the interview.

🏢 For companies: how to compete for talent

Now the contrary. If you’re a hiring manager or founder reading this, you already know the challenge. You’re getting hundreds of applications per role and still can’t find the right people. Startups are competing with Isomorphic Labs (£1.6 billion raise in 2025), Google DeepMind, and Boltz for the same small talent pool. Joe’s take on what works:

“A lot of candidates want more ownership, closer access to founders, more influence over direction of a product, and thrive on the ability to actually see their work shape a product directly. The larger players cannot offer this at the same level.”

Smaller companies can lean into that. What the ones that hire well do differently:

• Clarity of mission: Strong conviction about what they are building and why it matters. “If people believe in the founding team and what they’re standing for, it counts for a lot.”

• Process efficiency: “Slow feedback, too many stages, or technical tasks that take hours of a candidate’s time can quickly put people off.” Even unsuccessful candidates should leave with a good impression.

• Communication throughout: “Companies sometimes assume that if there’s no update, there’s no reason to contact the candidate. From the candidate’s side, silence feels like being ghosted.” Even an update saying they are still in consideration keeps people engaged.

• Clear expectations: “So often role details change mid-search, sometimes several times, which sends a mixed message to market and damages perception to the target talent pools.”

What we think: The talent competition in bio-AI is asymmetric in a way that favours startups, if they play it right. The big players offer prestige and salary. They cannot offer speed, ownership, or the feeling of shaping something from scratch. The startups that lose candidates to Isomorphic or DeepMind are usually the ones that ran a slow, unclear process, not the ones that lost on compensation alone. Your hiring process is your first product demo. If it’s confusing or inconsistent, strong candidates will read that as a signal about what working there is actually like.

⚡ The market in motion

Two things are true at the same time in computational biology right now. Traditional pharma is contracting (patent cliffs, layoffs, restructuring), while AI-native biotech is expanding fast. Isomorphic Labs raised £1.6 billion with no molecules in clinical trials. UK seed investment leapt 19% in 2025. Two UK biotech companies hit unicorn status (Verdiva Bio and Isomorphic Labs). The computational biology market overall is growing at 13% annually toward $22 billion by 2034.

For candidates, the opportunity is real, but “learn Python and apply broadly” doesn’t work anymore. The market rewards specialists who can ship production systems and communicate across disciplines. Infrastructure roles (MLOps, GPU optimisation, deployment) are where the most acute demand is. The salary premium for ML over traditional bioinformatics is widening. Remote work exists, though it’s not the default.

The people who do best in this market aren’t necessarily the most credentialed. They tend to be the ones who are visible, who adapt quickly, and who build things in the open.

💬 Want to be featured in Kiin Bio Weekly?

Each issue we speak directly with researchers, scientists, and builders working at the frontier of AI in life sciences. If you're working on something in this space and think it would resonate with our community, I'd love to hear from you. Fill out this form or reach out to me directly.

Found this useful? Forward it to a colleague in computational biology, it's the best way to help the newsletter grow.

Share Kiin Bio Weekly

Subscribe now to stay at the forefront of AI in Life Science. Every week: primers, deep dives, and direct conversations with the people building the field.

Subscribe now

Connect With Us

Have questions on this or suggestions for our next deep dive? We’d love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

Three papers this week that all ask some version of the same question: are we actually measuring the right thing? SwitchCraft pushes protein design past the single-structure assumption that underpins most current methods. TadA-Bench asks whether protein language models can predict the future of evolution or only interpolate its past. Chem-PerturBridge harmonises the messy pile of perturbation transcriptomics data and reveals just how little of it agrees at the gene level. The connecting thread: the foundation models exist, but the design paradigms and benchmarks haven’t caught up.

We just opened up our Kiin Pioneer Programme: free access to our platform for academic and nonprofit research teams for a year.

The short version: we’ve built a place where scientists can collaborate on their drug discovery work without everything living in disconnected tools and someone’s local files. Literature reviews, target discovery, bioinformatics, all in one place. When one person finds something interesting and someone else has relevant data, the platform catches that and suggests what to look at next. Everything’s tracked, so six months later you actually know what was done and why.

We’re looking for teams who are trying to move faster on questions like: which targets should we prioritise? How do we make sense of conflicting evidence? What’s actually worth testing next?

No cost, no data transfer, all IP stays with your institution. Applications close in August, first cohort starts in September.

Read more about the programme

Apply now

SwitchCraft: A Programmatic Framework for Designing State-Switching Proteins

🔬 Protein design tools optimise for one structure. Real proteins switch between conformations to do their jobs: allosteric enzymes toggle between active and inactive states, biosensors change shape upon binding. Designing proteins that intentionally switch between defined states has been a manual, low-throughput exercise.

Jing, Bafna, and colleagues from MIT’s Berger lab built SwitchCraft, a framework that designs proteins with specified multi-state behaviour by backpropagating through differentiable structure prediction models.

🧬 The core idea: treat Boltz-1 (an open-source structure prediction model) as a differentiable loss function. Define the desired structural states, then optimise the sequence so that it folds into all of them under appropriate conditions. The framework is programmatic. You compose design objectives from modular building blocks rather than training a new model per task.

⚡ They demonstrate allosteric regulation of protein motifs, discrimination between bound ligand identities, and fluorescent biosensor design. The biosensor results are particularly telling: designed sequences show distinct fluorescence states depending on which ligand is bound. That’s functional multi-state behaviour that hasn’t been accessible through standard single-state design pipelines.

🧪 Where This Fits

This sits at the leading edge of computational protein design, downstream of structure prediction but upstream of experimental characterisation. The reason it exists now is simple: differentiable structure prediction. Until models like Boltz-1, ESMFold, and AlphaFold became fast and accurate enough to use as gradient-providing modules, you couldn’t backpropagate through structure. RFdiffusion and ProteinMPNN design single structures beautifully, but they don’t handle the multi-state problem. SwitchCraft does something conceptually different: it treats structure prediction as a subroutine rather than the end goal. The limitation is that experimental validation here is still computational, relying on Boltz-1’s predictions of the designed states. Whether these sequences actually fold into multiple states in a wet lab remains open, though the biosensor designs are testable. If you work on biosensors, allosteric switches, or molecular logic gates, this is worth trying now.

💡 Why This Is Cool

This is what it looks like when structure prediction becomes infrastructure rather than the main event. The field spent five years building accurate folding models. Now those models are components in design loops. Multi-state protein design has been a goal since the Kuhlman lab’s early work on conformational switches, but the computational tools never matched the ambition. SwitchCraft doesn’t solve the full problem (experimental validation is still the bottleneck), but it makes the design step tractable in a way it simply wasn’t before.

📄 Read the paper.

💻 Try the code.

TadA-Bench: A Million-Variant Benchmark for Future-Round Discovery Toward Agentic Protein Engineering

🔬 Protein fitness prediction benchmarks typically test whether models can fill in gaps within a mutational landscape. That’s interpolation. What protein engineers actually need is extrapolation: given rounds 1 through 10 of directed evolution, can you predict what we’ll find useful in round 11? No existing benchmark properly tests this.

Gao and colleagues from Shanghai Jiao Tong University built TadA-Bench from 31 rounds of real TadA (tRNA adenosine deaminase) directed evolution, totalling roughly one million variants.

🧬 The benchmark enforces chronological evaluation: models train on earlier rounds and must predict which variants from later rounds will be experimentally validated. It provides aligned DNA, RNA, and protein sequences, and uses a Seq2Graph method to create comparable activity measurements across rounds that originally used different assay conditions.

⚡ The headline finding is sobering. Current protein language models (including ESM-2 and various fine-tuned variants) struggle with temporal prediction even when they perform well on standard interpolation benchmarks. The gap between interpolation and extrapolation performance is substantial. Good performance on DMS datasets doesn’t mean your model can guide the next round of experiments. Many practitioners suspected this; now there’s a number attached to it.

🧪 Where This Fits

This is a benchmark, not a tool, and its value depends on adoption. It fills a gap the field has known about for years. ProteinGym and similar resources test interpolation well, but nobody had assembled a large-scale temporal benchmark from real directed evolution campaigns. The 31 rounds of TadA evolution provide unusually rich temporal structure (most published datasets have 3-5 rounds at most). The “agentic protein engineering” framing is timely: as more groups wire LLMs into experimental design loops, you need evaluation frameworks that test whether the model’s suggestions actually lead somewhere productive. The dataset is on Hugging Face and the code is open on GitHub, which lowers the barrier to adoption. The main caveat is generalisability: TadA is one enzyme. Performance on this benchmark won’t guarantee performance on your protein of interest. Still, it’s the best temporal test we have.

💡 Why This Is Cool

The protein ML field has a measurement problem. Papers report NDCG on held-out DMS positions and claim their model “guides protein engineering.” This benchmark calls that bluff. Can your model predict the future, or only reconstruct the past? The distinction matters enormously for anyone running a real directed evolution campaign. If current models fail at temporal prediction (and they largely do), that’s uncomfortable but useful information. It tells you where the actual research gap is, which is more valuable than another leaderboard.

📄 Read the paper.

💻 Code and data.

🤗 Access the dataset.

Chem-PerturBridge: A Harmonized Compendium of Small Molecule Perturbation Transcriptomic Effects

🔬 The field has generated enormous amounts of small molecule perturbation transcriptomics data (L1000, sci-Plex, Tahoe, and many smaller datasets). The problem: nobody knows how well they agree with each other, and combining them for model training requires harmonisation that hasn’t been done systematically.

Szałata and colleagues from the Theis lab at Helmholtz Munich built Chem-PerturBridge, standardising 37,000+ compounds across 1.25 million samples from eight assay types with consistent metadata.

🧬 The compendium spans nine datasets including sci-Plex3, Tahoe, L1000, OP3, DILImap, and others. The harmonisation pipeline normalises cell line annotations, compound identifiers, and gene nomenclature, making cross-dataset comparison possible for the first time at this scale.

⚡ The uncomfortable finding: gene-level agreement between datasets is poor. When the same compound is tested in the same cell line across different platforms, correlation at the individual gene level is low. Directional consistency (is the gene up or down?) is better, and turns out to be sufficient for improving compound representation learning. Models trained on the harmonised directional data outperform those trained on any single dataset alone

🧪 Where This Fits

This sits upstream of almost everything in computational chemical biology: virtual screening, target identification, mechanism-of-action prediction, and toxicity forecasting all depend on perturbation data quality. The honest finding here is that the perturbation transcriptomics field has a reproducibility problem at the gene level, even between high-quality experiments. That’s not new as a suspicion, but quantifying it across eight assay types is valuable. The positive takeaway is that directional signals persist, and they’re enough to learn useful compound representations. For practitioners, this means you should probably stop training on raw gene-level expression values from a single perturbation study and start using directional features from multiple sources. The resource is open (MIT licence on code, upstream licences on data), which is appropriate for something positioned as community infrastructure.

💡 Why This Is Cool

The field has been treating perturbation transcriptomics datasets as interchangeable training data without checking whether they actually say the same thing. They don’t, at least not at the resolution most people assume. This paper does the unglamorous work of quantifying that disagreement and showing what signal does survive. The practical consequence: if you’re building models on LINCS L1000 data alone, you’re probably leaving performance on the table. Directional agreement across platforms is a more robust training signal than absolute expression values from one platform. This is infrastructure work rather than a methods advance, but it’s the kind that makes everything downstream more trustworthy.

📄 Read the paper.

💻Try the code.

🗓️ Events & Competitions

The best competitions, hackathons, and community challenges in AI x life sciences, curated weekly. Know something worth featuring? Reply and let us know.

More upcoming events:

Creative Disruption Forum: Modern Drug Discovery | June 18, NIAB Cambridge

A full-day forum for biotech and R&D leaders exploring how technology is changing small molecule drug discovery. Keynote interviews with industry thought leaders followed by workshops under Chatham House Rules, limited to 60 attendees. Part of Cambridge Wide Open Week. Organised by Graham Combe and Prof Tony Sedgwick. £60 for biotech companies.

London Protein Design Day | June 23, Imperial College London

The first edition of a one-day symposium bringing together London’s protein design community and beyond. Programme spans AI-driven design, molecular dynamics, and bioinformatics, with applications across enzymes, antibodies, and materials. Organised by Pietro Sormanni, Rebecca Birolo, and Jakub Lála. Abstract deadline for poster/oral presentations is this Saturday (May 17). In person only.

BioHackathon Europe 2026 | November 9-13, Barcelona

ELIXIR’s annual international bioinformatics hackathon, running since 2018. 160+ participants, five days of collaborative coding on open bioinformatics infrastructure and tools. The call for project proposals opens March 16 and closes April 15 - so if you want to lead a project, that’s your window.

Thanks for reading!

💬 Get involved

We’re always looking to grow our community. If you’d like to get involved, contribute ideas or share something you’re building, fill out this form or reach out to me directly.

Connect With Us

Have questions or suggestions? We'd love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

I keep coming back to this question of whether the protein ML field has a modelling problem or a data problem. This week's papers make the case from both sides. Biohub's ESM release argues that scale and architecture can solve binder design from scratch. OpenADMET argues the opposite: that no model will work until the training data stops being inconsistent rubbish. SynFit sits somewhere in between, showing what you can get when you have good multi-property data and a framework that knows how to use it.

We just launched our Kiin Pioneer Programme, giving academic and nonprofit research teams one year of free access to our drug discovery platform!

KiinOS is a platform where scientists can run literature reviews, target discovery, and bioinformatics in one place. It keeps a record of what’s been done, by who, and what came out of it. So if one person finds a promising target and someone else has relevant data, the platform connects those results and suggests what to pursue next.

We’re looking for teams asking: which targets should we prioritise? How do we interpret conflicting evidence? Which hypotheses are worth testing next?

There’s no cost, no data transfer, and all IP stays with your institution. Applications close in August, with the first cohort starting September.

Read more about the programme

Apply now

ESM: A World Model of Protein Biology

🔬 We can predict how proteins fold, but designing new ones that actually bind specific targets, particularly antibodies, still requires months of expensive lab screening just to find starting candidates.

Biohub has released ESM: a protein language model (ESMC, trained on 2.8 billion sequences), a structure prediction and design model (ESMFold2), and a map of 6.8 billion sequences (ESM Atlas). All MIT-licensed.

🧬 ESMFold2 learns protein representations from evolutionary data, then searches that learned space for proteins predicted to bind a given target. It scores candidates using its own confidence estimates, so the entire design loop is computational. Structure prediction runs from a single sequence without needing alignment databases.

⚡ Designed binders for five cancer/immunology targets were validated in the lab: minibinder success rates of 70%, scFv antibody success rates of 21%. A PD-L1 binder hit 4.3 nM affinity and blocked immune checkpoint suppression in cells. Cryo-EM confirmed an EGFR binder matched the prediction at 1.2 Å RMSD.

🧪 Where This Fits

This positions ESM as a competitor to both AlphaFold 3 (for structure prediction) and RFdiffusion/ProteinMPNN (for binder design). The antibody-antigen prediction results are interesting because this is where AlphaFold has historically struggled most. The binder design piece is where Biohub is making their boldest claim: that the earliest stage of therapeutic protein discovery can happen computationally in days rather than months. Worth noting that the 21% scFv success rate, while a step up from near-zero for most computational methods, still means roughly 80% of designs fail in the lab. These are first-round hits. David Baker’s group (RFdiffusion) and Generate Biomedicines (Chroma) are the obvious comparisons for generative protein design. ESMFold2 is differentiated by working from a language model backbone rather than a diffusion architecture, which changes how it scales with compute at inference time.

💡 Why This Is Cool

The sparse autoencoder analysis is what I find most thought-provoking. ESMC independently recovered biological concepts like the nucleophilic elbow motif across 75 of 99 relevant enzymes, despite never being told what one is. That’s a model learning the grammar of protein biology from sequence alone. Whether the binder design numbers hold up across a broader and more difficult target set remains to be seen. The cryo-EM validation is reassuring, but five targets is still five targets.

📄 Read their press release.

💻 Try the tool.

SynFit: Synergistic Contrastive Learning for Multi-Objective Protein Fitness Prediction and Optimisation

🔬 Protein engineering almost always requires optimising multiple properties at once, but current ML fitness predictors handle each property independently. Train separate models for yield and selectivity and you’ll get variants that excel at one while tanking the other.

Georgia Tech and UC Santa Barbara developed SynFit, a multi-objective framework that fine-tunes protein language models on experimental fitness data across multiple assays simultaneously.

🧬 SynFit combines a shared ESM2 encoder with property-specific prediction heads, using contrastive learning to capture cross-property relationships from deep mutational scanning data. Predictions are integrated via Pareto sorting to find variants that improve everything at once.

⚡ On Pareto front analysis across 20 proteins, SynFit hits the optimal front 70% of the time versus 60% for ProteinNPT and 55% for ConFit. The wet-lab result is more convincing: 83 out of 100 designed hextuple mutants for a biocatalytic borylation enzyme showed simultaneously improved yield and enantioselectivity, with multiple variants beating everything in the training data.

🧪 Where This Fits

This sits squarely in the protein engineering workflow after you have initial DMS data and want to explore combinatorial sequence space. ConFit, from the same group, is the direct predecessor, and SynFit extends it with the multi-objective training component. ProteinNPT (Notin et al., ICML 2023) tackles a related problem through non-parametric transformers but doesn’t explicitly model cross-property correlations. The wet-lab validation on biocatalytic borylation is well chosen because it’s a genuinely new-to-nature reaction where directed evolution data is sparse, making the Pareto optimisation problem harder. The limitation is that you still need initial multi-property DMS data for your protein of interest, which isn’t always available. The KRAS case study (identifying shared functional residues across six binding partners) is a nice mechanistic demonstration but relies on an unusually well-characterised system.

💡 Why This Is Cool

The 83/100 result is more impressive than any of the benchmark numbers. Getting computationally designed hextuple mutants to simultaneously beat the training set on both yield and enantioselectivity, in a single round without iterative screening, is a practical result that enzyme engineers will care about. It suggests ML-guided combinatorial design can start to compress the “design-build-test” cycle for multi-objective problems. The architecture is straightforward enough that adoption shouldn’t be difficult once the code is released.

📄 Read the paper.

💻Try the code.

Mapping the Avoid-ome: A Systematic Open-Science Approach to Predictive ADMET

🔬 Around 30% of clinical drug failures trace back to ADMET problems. The ~100 proteins responsible (CYPs, hERG, transporters, nuclear receptors) are well known, but existing ML models train on data cobbled from dozens of labs using different protocols. A recent analysis found almost no correlation between IC50 values for the same compound measured by different groups.

Fraser (UCSF), Edgar (Octant), Chodera (MSKCC), and Walters (OMSF) have launched OpenADMET, an ARPA-H and Gates Foundation-funded consortium generating systematic, internally consistent ADMET datasets and releasing everything publicly.

🧬 The consortium runs assays across the full “Avoid-ome” panel at industrial scale: 30,000 compounds per run in 1536-well plates, under $0.40 per compound. Active learning selects informative compounds for expansion, and structural biology (100+ PXR crystal structures so far) resolves binding modes.

⚡ This is a programme-level perspective paper rather than a single dataset release. The first community challenge has run. They’re screening tens of thousands of compounds weekly. I appreciate the honesty that this is long-term infrastructure. They’re not claiming to have solved ADMET prediction; they’re arguing nobody will until the data problem is addressed.

🧪 Where This Fits

This is an upstream data-generation effort, not a prediction tool. It sits before everything else in the ADMET pipeline: the datasets it produces will train the next generation of models. TDC (Therapeutics Data Commons) and ChEMBL aggregate existing literature data. OpenADMET’s bet is that literature data is too inconsistent to train reliable models, and that generating internally consistent measurements from scratch, with structural validation, justifies the cost. Tools like ADMET-AI (Swanson et al., 2024) would be downstream consumers of these datasets. The federated learning alternative (training behind pharma company firewalls) is explicitly discussed and dismissed: it can’t generalise beyond local chemical space and doesn’t produce the structural understanding needed for true mechanistic models.

💡 Why This Is Cool

The framing is what matters here. By defining ADMET as a finite structural biology problem (map the interactions with roughly 100 proteins and you’ve covered most failure modes), they turn an open-ended prediction challenge into a bounded experimental campaign. The question is whether $0.40-per-compound assays and active learning can produce models that generalise to novel chemical matter outside the training distribution. The open-science commitment, with Gates Foundation backing, makes this more credible than most “we’ll share data eventually” promises from pharma-adjacent initiatives.

📄 Read the paper

💻 Learn more

🗓️ Events & Competitions

The best competitions, hackathons, and community challenges in AI x life sciences, curated weekly. Know something worth featuring? Reply and let us know.

More upcoming events:

Creative Disruption Forum: Modern Drug Discovery | June 18, NIAB Cambridge

A full-day forum for biotech and R&D leaders exploring how technology is changing small molecule drug discovery. Keynote interviews with industry thought leaders followed by workshops under Chatham House Rules, limited to 60 attendees. Part of Cambridge Wide Open Week. Organised by Graham Combe and Prof Tony Sedgwick. £60 for biotech companies.

London Protein Design Day | June 23, Imperial College London

The first edition of a one-day symposium bringing together London’s protein design community and beyond. Programme spans AI-driven design, molecular dynamics, and bioinformatics, with applications across enzymes, antibodies, and materials. Organised by Pietro Sormanni, Rebecca Birolo, and Jakub Lála. Abstract deadline for poster/oral presentations is this Saturday (May 17). In person only.

BioHackathon Europe 2026 | November 9-13, Barcelona

ELIXIR’s annual international bioinformatics hackathon, running since 2018. 160+ participants, five days of collaborative coding on open bioinformatics infrastructure and tools. The call for project proposals opens March 16 and closes April 15 - so if you want to lead a project, that’s your window.

Thanks for reading!

💬 Get involved

We’re always looking to grow our community. If you’d like to get involved, contribute ideas or share something you’re building, fill out this form or reach out to me directly.

Connect With Us

Have questions or suggestions? We'd love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

This week we’re looking at clinical AI, specifically the type that nobody’s talking about. I got connected to Ignacio through his work at Savana, where they’ve been extracting real-world evidence from clinical records for years. What struck me the most was how much of the conversation around AI in medicine misses the point: everyone’s focused on ChatGPT and scribes, while the predictive models that actually enable personalised medicine are sitting there with regulatory approval and population-level evidence, largely unused.

We got on a call, and the result is this primer.

Freebie alert: We know how hard science is. That’s why we built the Pioneer Programme.

We’re selecting academic and nonprofit research teams to get one year of free access to our drug discovery platform plus hands-on support from our science team. If your research bottleneck isn’t data but connecting the findings you already have, this is for you.

We’re looking for teams asking: which targets should we prioritise? How do we interpret conflicting evidence across datasets? Which hypotheses are worth testing next? Where are the strongest translational opportunities?

No cost. No data transfer. All IP stays with your institution. Applications close August, cohort starts September.

Read more about the programme

Apply now

Healthcare systems across Europe are breaking. Waiting lists stretch to 14 months. The UK Health Secretary declared the NHS “broken.” Spain is close behind. Populations are ageing, medicines are expensive, and the workforce cannot scale. Into this crisis arrives artificial intelligence, not as a future promise, but as a present reality. The FDA has approved close to 1,500 AI use cases across clinical specialties. Meta-analyses now demonstrate cost-effectiveness in diabetes, colon cancer, and mammography screening. The question is no longer whether AI works in medicine. It is why adoption remains so slow.

Yet most clinicians, when asked about AI, will talk about ChatGPT and medical scribes. They are looking at the louder revolution while the quieter, more consequential one unfolds beneath it.

We spoke to Ignacio H. Medrano, neurologist-turned-CEO of Savana, about the real state of clinical AI: what is proven, what is hype, and what clinicians are getting wrong about the pace of change.

Ignacio H. Medrano, CEO of Savana

“It’s unethical not to use AI these days in certain areas, because it’s proven, and it’s proven at scale across populations.”

🔀 Two types of AI, one common misunderstanding

The most pervasive misconception among clinicians today is that generative AI (LLMs, chatbots, multi-modal models) is the only AI that matters. It is visible, fast-moving, and immediately useful: scribes that eliminate documentation burden, literature tools like Open Evidence replacing PubMed searches, agents managing waiting lists. These applications are exploding because they save time and, critically, do not require clinical validation. They handle documentation, not decisions.

But the deeper disruption is discriminative AI: predictive, classification-based models that have existed for over a decade. This is the AI that enables precision medicine. Granular predictions for individual patients about immunotherapy response, relapse probability in multiple sclerosis, optimal drug sequencing in haematologic cancers. It takes statistics to a level where you can determine the actual probability of a specific outcome for a specific patient.

Discriminative AI started earlier but arrives later in practice. Every algorithm requires validation, external replication, meta-analysis, and integration into clinical guidelines. That pipeline is slow. But it is the pipeline that delivers personalised medicine, and it is now producing results.

“A big misconception is forgetting that discriminative, predictive AI is the real silent disruption,” says Medrano. “The other is underestimating the speed at which agentic AI is arriving. People think this takes 20 years. It’s happening now, like thunder.”

✅ Separating signal from noise: what makes clinical AI “real”

With close to 1,500 FDA-approved AI applications and an exponential curve of machine learning publications on PubMed, distinguishing proven tools from hype requires a framework.

Medrano uses three levels:

1. Regulatory approval: FDA or EMA clearance confirms correct dataset construction, generalisation to new cohorts, and absence of bias. This is the minimum threshold.

2. Population-level evidence: Publications demonstrating that algorithms work in general populations and are cost-effective. Meta-analyses now exist for AI in diabetes management, colon cancer screening, and mammography. A National Library of Medicine study proved chatbots reduce severe mental health events. A cardiology trial demonstrated that AI applied to ECGs reduces cardiovascular mortality.

3. Real-world deployment: Hospitals actually running these systems in clinical workflows. Hospital Clinic Barcelona has used AI to predict sepsis in intensive care for over three years. Finland deploys predictive models on GP workstations for population segmentation. In China, Ping An Good Doctor attends 100 patients daily without human involvement. In Utah, AI autonomously renews prescriptions.

The gap between levels two and three, between proven effectiveness and actual deployment, is where the real problem lives.

🚧 Three barriers, one that matters most

Technical: Clinical information remains fragmented across systems. Standards like HL7 FHIR are improving interoperability, but the problem is not fully solved. Training and validating models still requires stitching together disparate data sources.

Regulatory: Largely resolved. European regulation now recognises that anonymised data used for research does not require individual informed consent. The European Health Data Space will mandate hospitals to share data within one to two years. Countries like Switzerland, France, Germany, and the UK already permit compliant secondary use of clinical data.

Cultural: The real bottleneck. Not rejection; most managers and clinicians accept AI is inevitable. The problem is twofold. First, pure ignorance: hospital managers do not realise they could build pharmacogenomic models today that predict which patients will respond to expensive biologics, saving millions by paying only for drugs that will work. Second, politics: an algorithm that reduces the need for human labour is nearly impossible to sell when unions and the public demand 10 new nurses before any technology investment. Politicians choose nurses even when they understand AI’s value.

“It’s not belief anymore, it’s knowledge,” says Medrano. “Managers just don’t understand the power of what’s already possible with the data they have.”

📊 Real-world evidence: from luxury to necessity

Real-world evidence has always mattered. It is the difference between reading about a country and landing there. Clinical trials tell you what should happen under controlled conditions. Real-world evidence tells you what actually happens.

The barrier was always collection. Patient by patient, variable by variable, manually assembling registries over years. Exhausting, expensive, and therefore underutilised. Now, computational systems can extract this information automatically, reliably, and at scale from electronic health records. Once extraction became feasible, demand exploded. Regulators began requesting it. Pharma began requiring it.

This is where initiatives like the UK’s Foresight programme become significant: 57 million medical records feeding predictive models for 100 diseases at 20-year horizons. The Scandinavian countries (Norway and Denmark) are sharing data internationally and validating models across borders. These are not pilots. They are national-scale infrastructure decisions.

🤖 The convergence: agentic AI meets predictive models

Here is where the field is heading, and where most clinicians have not yet looked. The users of sophisticated discriminative AI models (multi-modal predictive algorithms trained on clinical text, genomics, proteomics, radiomics) will not be human doctors. They will be certified agentic AI systems.

Generative AI agents will orchestrate clinical workflows. Discriminative AI will provide the predictions those agents act on. The agent decides what to ask; the predictive model provides the answer. The human clinician supervises, validates, and handles what requires physical presence.

A meta-analysis of 15 studies already shows that in 13 out of 15, AI chatbots were rated as more empathic than human clinicians. Not even communication, the last presumed advantage of human clinicians, remains unchallenged.

“If doctors keep doing the same thing, they’ll become pointless,” says Medrano. “People will turn to their phones, where an agentic AI healthcare service for 10 euros will give them advice, pull their data, run algorithms. And then that agent will hire humans to perform the physical tasks it cannot.”

⚡ Where this leaves us

The infrastructure is being built. Data-sharing mandates are arriving. Validation evidence is accumulating. The two types of AI, generative and discriminative, are converging toward agentic systems that will reshape how healthcare is delivered.

The bottleneck is not technology or regulation. It is the speed at which institutions recognise what is already possible, and act before the system breaks entirely, or before patients simply route around it.

Big thanks Ignacio for meeting with us and sharing his insights for this primer!

💬 Want to be featured in Kiin Bio Weekly?

Each issue we speak directly with researchers, scientists, and builders working at the frontier of AI in life sciences. If you're working on something in this space and think it would resonate with our community, I'd love to hear from you. Fill out this form or reach out to me directly.

Found this useful? Forward it to a colleague in drug discovery or computational chemistry, it's the best way to help the newsletter grow.

Share Kiin Bio Weekly

Subscribe now to stay at the forefront of AI in Life Science. Every week: primers, deep dives, and direct conversations with the people building the field.

Subscribe now

Connect With Us

Have questions on this or suggestions for our next deep dive? We’d love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

We know how hard science is. That’s why we built the Pioneer Programme.

We’re selecting academic and nonprofit research teams to get one year of free access to our drug discovery platform plus hands-on support from our science team. If your research bottleneck isn’t data but connecting the findings you already have, this is for you.

We’re looking for teams asking: which targets should we prioritise? How do we interpret conflicting evidence across datasets? Which hypotheses are worth testing next? Where are the strongest translational opportunities?

No cost. No data transfer. All IP stays with your institution. Applications close August, cohort starts September.

Read more about the programme

Apply now

Proteo-R1: Reasoning Foundation Models for De Novo Protein Design

🔬 Current deep learning methods for protein design generate molecular structures without explicitly reasoning about which residues matter or why. All residues are treated uniformly during generation, and design intent is implicitly buried in diffusion parameters. This makes it hard to interpret why a model made a particular choice, or to reuse that logic on a different target.

Researchers from Stanford, RIKEN, and collaborating institutions introduce Proteo-R1, a framework that separates molecular understanding from geometric generation. A multimodal large language model first reasons about binding interactions, then passes residue-level constraints to a diffusion model that generates the structure.

🧬 The understanding expert (a multimodal LLM) analyses protein sequences, AF3-style structural representations, and textual context to identify key interaction residues and predict their amino acid identities. These sparse, residue-level decisions are passed as hard constraints to the generation expert, an AlphaFold3-style diffusion model that performs conditional co-design while respecting the fixed interaction anchors. Training proceeds through three stages: multimodal alignment, structural reasoning mid-training, and joint reasoning-guided design on antibody-antigen complexes from SAbDab.

⚡ On simultaneous multi-CDR antibody redesign, Proteo-R1 achieves the lowest or near-lowest per-CDR RMSD in five of six regions, with interface improvement (IMP) of 56.58%. It produces the lowest steric clash rates (0.50% intra-chain, 0.14% inter-chain) and best dihedral distribution divergence among all tested methods. On CDR-H3 design via the RAbD benchmark, it obtains the best lDDT (0.9693), TM-score (0.9816), and DockQ (0.801) over DGENet, BoltzGen, and MFDesign.

🔬 Applications and Insights

1️⃣ Interpretable Antibody Engineering

The reasoning expert produces explicit, human-readable justifications for which residues it selects as interaction anchors. You can inspect and modify the design logic directly.

2️⃣ Controllable CDR Design

Researchers can edit the reasoning outputs (e.g. specifying different hotspot residues) to steer the generative model without retraining, giving fine-grained control over binding specificity.

3️⃣ Structure-Sequence Consistency

Proteo-R1 is the only method that achieves positive structure-sequence consistency (Δ = IF-AAR minus AAR) across five of six CDR regions. Its designs are structurally grounded rather than simply recovering native sequences.

4️⃣ Generative Backend Flexibility

The reasoning expert works with alternative generative frameworks. Tested with UniMoMo, it improves IMP from 65% to 67.79% and binding energy from 8.46 to 7.35 ΔG without architectural changes to either component.

💡 Why This Is Cool Proteo-R1 separates the “what should we build” question from the “how do we build it” question. That mirrors how human protein engineers actually work: identify critical interaction residues first, then optimise geometry under those constraints. The reasoning module plugs into different generative backends, and the design logic is readable and editable rather than locked inside a diffusion trajectory.

📄 Read the paper

💻 Try the code

ActivityFinder: Toward the Fully Automatic Integration of Structural and Binding Affinity Data

🔬 Building computational models that predict binding affinity requires both 3D protein-ligand structures and experimental activity measurements. These data live in separate databases (PDB for structures, ChEMBL for bioactivities) with no fully automated way to link them. Identifier-based approaches miss connections where sequences diverge, ligand representations vary, or binding-site mutations are present.

A team at the University of Hamburg developed ActivityFinder, a fully automated pipeline that links crystal structures of protein-ligand complexes directly to wet-lab bioactivity data. It requires only PDB files and a ChEMBL database dump, with no external services or continuous data connections.

🧬 ActivityFinder works in two stages. First, it builds an ActivityDB instance by parsing PDB structures into the NAOMI data format, extracting ligand representations as six different string types (InChI, InChIKey, canonical SMILES with and without stereo, InChI connection layer, InChI hydrogen layer), and creating BLAST databases from protein sequences. Second, it queries this database using sequence alignments (at 80%+ identity) and detailed chemical structure matching to cross-reference structural and bioactivity records. It tracks mutations and binding-site residues at atomic resolution.

⚡ Applied to 226,302 PDB structures and ChEMBL 35, ActivityFinder linked 20,197 PDB entries involving 13,734 PDB ligands to 17,829 unique ChEMBL ligands across 2,585 ChEMBL targets, covering over one million bioactivity data points. Compared to BioChemGraph (an identifier-based method), ActivityFinder identifies 46,287 unique structure-activity triplets versus 19,333. Of the 6,575 triplets unique to ActivityFinder, 1,082 are entirely novel: new combinations of PDB complex, ChEMBL target, and ChEMBL molecule.

🔬 Applications and Insights

1️⃣ Proprietary Data Integration

Because it requires only local PDB files and an SQL database, ActivityFinder works for pharmaceutical companies that want to link in-house crystal structures to internal bioactivity records without sending data to external services.

2️⃣ Mutation-Aware Linking

The tool explicitly tracks sequence variants and binding-site mutations at atomic resolution. Researchers can study how specific point mutations affect ligand binding affinity across related structures.

3️⃣ Training Data for ML Scoring Functions

The linked structure-activity pairs are ready-made training datasets for machine learning scoring functions. Quality annotations (confidence levels 1 to 5) let modellers filter for the precision their application requires.

4️⃣ Expanding Known Chemical Space

Of the novel triplets ActivityFinder uniquely identifies, many involve new combinations of PDB complex, ChEMBL target, and ChEMBL molecule not found by any other method. These are new data points for structure-based drug design.

💡 Why This Is Cool

Everyone building a scoring function or binding predictor eventually has to manually assemble their own structure-activity dataset. The results are rarely shared and identifier matching misses a lot. ActivityFinder automates this from scratch, works on proprietary data, and finds 2.4x more structure-activity links than the identifier-based alternative.

📄 Read the paper

💻 Try the tool: Available via the ProteinsPlus REST API

ProteinConformers: Large-Scale and Energetically Profiled Descriptions of Protein Conformational Landscapes

🔬 Understanding protein function requires capturing how structures move across their conformational space. Existing MD trajectory databases start only from native structures (near the global energy minimum), conformer generators have no standardised benchmarks, and available datasets provide limited energetic annotations. No resource maps the full spectrum from non-native to near-native states with both structural and energetic characterisation.

Researchers at the National University of Singapore (Yang Zhang’s group) present ProteinConformers, a database of 2.7 million geometry-optimised conformations across 734 proteins, paired with energy evaluations and a benchmarking framework for multi-conformation generators.

🧬 The dataset uses a multi-seed decoy sampling strategy: for each protein, hundreds of diverse starting conformations (drawn from CASP5-15 prediction submissions) are each run through full-atom molecular dynamics simulation using GROMACS 2023 with the OPLS-AA force field. Each conformation receives five energetic evaluations (RW, RWplus, EvoEF2, Rosetta, FoldX) and pairwise similarity annotations (TM-score and RMSD). The curated benchmark subset, ProteinConformers-lite, contains 381,546 MD-refined conformers across 87 CASP14/15 proteins with 1.9 million energetic annotations.

⚡ ProteinConformers spans protein lengths from 33 to 949 residues, with conformations distributed continuously from non-native to near-native states (TM-scores covering the full 0 to 1 range). Local geometric quality matches the Top2018 reference set: dihedral angle distributions show Pearson correlations of 0.97 to 0.99, and near-native conformations fall below the Top2018 average Ramachandran outlier rate of 13%. In benchmarking five generative models, BioEmu achieves the highest coverage under strict energy thresholds (5 kJ/mol), while AlphaFlow-MD scores comparably on the CGMSmah geometric plausibility metric. Total compute cost: approximately 40 million CPU hours.

🔬 Applications and Insights

1️⃣ Benchmarking Conformation Generators

ProteinConformers-lite is the first standardised evaluation framework for models like AlphaFlow, ESMFlow, and BioEmu. It measures both diversity (how much of the energy landscape a model covers) and plausibility (geometric realism) in one benchmark.

2️⃣ Allosteric Mechanism Studies

The continuous energy surfaces from non-native to native states let researchers study allosteric transitions and functionally relevant intermediates that single-seed MD simulations would miss.

3️⃣ Drug Discovery Applications

Energy-annotated conformational ensembles support ensemble docking workflows, where sampling multiple receptor states improves virtual screening hit rates compared to docking against a single crystal structure.

4️⃣ Evaluating Energetic Realism

With five energy functions evaluated per conformation, the dataset allows systematic comparison of how well different generators produce physically plausible low-energy structures versus merely diverse ones.

💡 Why This Is Cool

If you want to compare AlphaFlow against BioEmu against ESMFlow, you currently have no standard reference to test against. ProteinConformers fills that gap. Every conformation has known energy and measured structural similarity to the native state, and the web platform lets you explore without running simulations.

📄 Read the paper

💻 Explore the database

🗓️ Events & Competitions

The best competitions, hackathons, and community challenges in AI x life sciences, curated weekly. Know something worth featuring? Reply and let us know.

More upcoming events:

London Protein Design Day | June 23, Imperial College London

The first edition of a one-day symposium bringing together London’s protein design community and beyond. Programme spans AI-driven design, molecular dynamics, and bioinformatics, with applications across enzymes, antibodies, and materials. Organised by Pietro Sormanni, Rebecca Birolo, and Jakub Lála. Abstract deadline for poster/oral presentations is this Saturday (May 17). In person only.

BioHackathon Europe 2026 | November 9-13, Barcelona

ELIXIR’s annual international bioinformatics hackathon, running since 2018. 160+ participants, five days of collaborative coding on open bioinformatics infrastructure and tools. The call for project proposals opens March 16 and closes April 15 - so if you want to lead a project, that’s your window.

Thanks for reading!

💬 Get involved

We’re always looking to grow our community. If you’d like to get involved, contribute ideas or share something you’re building, fill out this form or reach out to me directly.

Connect With Us

Have questions or suggestions? We'd love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

We know how hard science is. That’s why we built the Pioneer Programme.

We’re selecting academic and nonprofit research teams to get one year of free access to our drug discovery platform plus hands-on support from our science team. If your research bottleneck isn’t data but connecting the findings you already have, this is for you.

No cost. No data transfer. All IP stays with your institution. Applications close August, cohort starts September.

Read more about the programme

Apply now

Single-cell biology has spent the last few years borrowing ideas from language models. Several groups have adapted transformer architectures to gene expression data, including scGPT and Geneformer, both of which pretrain on massive single-cell datasets using masked modeling objectives.

Transformers. Pretraining. Big datasets. It all felt like progress. But if you ask most practitioners what they actually use day to day, the answer is still pretty modest: embeddings that work, transfer across datasets, and do not collapse when the technology changes.

That tension is what motivated scConcept, a new foundation model for single-cell transcriptomics that steps away from gene reconstruction and instead asks a simpler, more explicit question: do these genes come from the same cell?

We spoke with Mojtaba Bahrami, in Fabian Theis’ lab at Helmholtz Munich, part of the the team behind scConcept about why the field needed a reset, what contrastive learning brings to biology, and why the future may be less about bigger models and more about better representations

🔴 The Problem

If you zoom out, most single-cell foundation models share the same basic idea. Treat genes like words. Mask some of them. Ask the model to predict what’s missing. This strategy mirrors masked language modelling introduced in BERT, where models learn by predicting missing tokens in a sentence. It works well enough on paper, but something feels off once you start using the embeddings downstream.

As Mojtaba Bahrami put it when we spoke, “In language models, training and inference are basically the same task. You predict the next token. But in single-cell, no one actually cares about predicting masked genes. People care about the embedding.”

That mismatch matters. Masked gene prediction optimises the wrong thing. The model gets good at reconstructing counts, but the cell-level representation becomes an afterthought. Most approaches average learned gene embeddings and hope for the best. This may work depending on the downstream question but often it doesn’t. If the question can not be simply answered by looking at gene level information and needs understanding higher level biological processes going on in the cell, then we have a problem.

The cracks show up quickly. Simple methods like PCA or VAEs still outperform large models on tasks like cell type annotation. Benchmarking efforts such as the scIB framework have shown that classical integration methods can remain highly competitive across datasets. Spatial assays break embeddings entirely. New technologies shift the latent space so much that “foundation” starts to feel like marketing rather than reality.

💡 The Idea

Here comes scConcept. Instead of asking the model to reconstruct genes, it asks something more aligned with how the actual biology is: What is the identity of a cell given a partial set of its gene expression. In other words, can we identify the biological processes that make up the identity of a cell consistently through looking at different views of its transcriptome profile?

Technically, this is done using contrastive learning. Each cell is split into two disjoint gene subsets. The model sees both views and is trained to pull them together in embedding space, while pushing apart views from different cells.

To make this work, the team borrowed a trick from BERT: a dedicated CLS token. In language, CLS represents a sentence. In scConcept, it represents the entire cell. The loss is applied directly to that token.

“If the model can solve this task,” Mojtaba explained, “it has to develop a high-level idea of cell identity. There’s no shortcut.”

That framing turns the embedding from a side effect into the main objective. The model is no longer rewarded for local gene accuracy, but for capturing global cellular identity.

📊 Where the data work really shows

The architecture alone is only half the story. The other half lives in how the data is presented. One of the team’s biggest insights came from a failed experiment. Early versions of scConcept used the same binning strategies as other models. The result was large, technology-driven shifts in the embedding space.

“That was the moment we realised something was fundamentally wrong,” Mojtaba said. “The model was learning the technology, not the biology.”

The fix was surprisingly simple. Instead of feeding absolute expression values, scConcept uses rank encoding. Genes are ordered by expression within each cell. Only relative relationships matter.

If gene A is higher than gene B, the model sees that. The actual counts are ignored.

This turns out to be remarkably robust. Different assays may disagree on absolute numbers, but gene rankings tend to stay stable. Rank encoding strips away much of the batch effect before the model even starts learning.

Then comes gene subsetting. During training, the model constantly sees partial views of cells, including realistic gene panels from spatial technologies. This forces the embedding to stay stable even when most genes are missing.

The result is a representation that does not panic when faced with a 300-gene panel instead of a full transcriptome.

📢 Why it is different

The evaluation results reflect that design choice. scConcept consistently outperforms other foundation models on cell type annotation, cross-technology transfer, and spatial imputation, often matching or beating domain-specific tools. More importantly, it fails more gracefully.

When information is missing, performance drops for the right reasons, not because the embedding space collapses. Closely related cell types remain close. Spatial structure is preserved. Adaptation improves things further without retraining the entire model.

One subtle but important point the authors stress is that scConcept is not a batch correction method. It does not erase real biological differences. It just stops the model from confusing technology with biology.

That distinction matters if you actually want to interpret what the model is doing.

🔮 The Future

The long-term vision goes further than single cells. Right now, scConcept is trained on around 30 million cells, deliberately matching the scale of existing models. The next step is obvious: train on hundreds of millions. Initiatives like the Human Cell Atlas and repositories such as CELLxGENE now host hundreds of millions of publicly available cells, making this scale increasingly realistic.

However, the more interesting shift is conceptual. As Mojtaba put it, “Cells are the starting point. But biology doesn’t stop there. Tissues matter. Spatial context matters. Patients matter.”

Contrastive learning over partial views opens the door to representing tissue sections, neighbourhoods, even whole samples. Instead of asking whether two gene sets come from the same cell, future models might ask whether two regions come from the same tissue state.

That feels like a natural progression, especially as spatial assays become routine.

For now, scConcept is a reminder that better questions often beat bigger models. By focusing on what practitioners actually use, rather than what looks impressive on paper, it points toward a quieter but more useful future for single-cell AI, which is probably what the field needs right now.

👨‍🔬 Get in touch with Mojtaba.

📑 Read the paper.

💻 Check out the code.

Thanks for reading Kiin Bio Weekly!

💬 Get involved

We’re always looking to grow our community. If you’d like to get involved, contribute ideas or share something you’re building, fill out this form or reach out to me directly.

Share Kiin Bio Weekly

Subscribe now to stay at the forefront of AI in Life Science and keep up with this upcoming season of deep dives.

Connect With Us

Have questions on this or suggestions for our next deep dive? We’d love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

Keeping up with AI x life science news can get exhausting.

It’s scattered across LinkedIn, X, Substack, arXiv, Slack, newsletters... and you still somehow miss the things that actually matter. Too much noise, not enough signal.

We’re building something to fix that: a smarter, more powerful way to stay on top of what’s actually relevant to you.

But we want to build it with you, not just for you. Take 2 minutes to tell us what’s missing. What you share will directly shape what we build, and you’ll be the first to benefit from it.

Share your input

🇺🇸 We’re heading to Bio-IT World in Boston, May 19-21.

Our CEO Filippo and CTO Bogdan will be there and would love to meet anyone thinking about:

• How AI is actually changing preclinical workflows (not just the hype)

• Why drug discovery is a systems problem, not just a science one

• What it takes to go from 5-year timelines to something radically faster

No pitch, just good conversation. If any of that’s on your mind, reach out - we’ll find a time to grab a coffee.

RegVelo: Gene-Regulatory-Informed Dynamics of Single Cells

🔬 RNA velocity models cellular dynamics but ignores gene regulatory interactions. Conversely, gene regulatory network inference methods neglect dynamics entirely. No existing approach jointly captures both, limiting our ability to simulate perturbations and predict how regulatory changes drive cell fate decisions.

RegVelo from Helmholtz Munich and Fabian Theis’s lab bridges this gap. It is an end-to-end deep generative framework that jointly infers transcriptome-wide splicing kinetics and gene regulatory interactions from scRNA-seq data, producing an actionable in silico cell for perturbation simulation.

🧬 RegVelo encodes unspliced and spliced RNA into a latent space, then models transcription as a regulated process governed by a GRN weight matrix. A parallel high-dimensional ODE solver couples all gene dynamics simultaneously rather than treating genes independently. Prior GRN knowledge from ATAC-seq or public databases constrains the network, while data-driven refinement learns new regulatory edges and edge weights.

⚡ On cell cycle data, RegVelo achieves a cross-boundary correctness of 0.864, velocity consistency of 0.873, and significantly outperforms scVelo and veloVI (p < 0.001). For GRN inference, it ranks first among six methods on edge prediction (median AUROC = 0.59) and achieves AUROC = 0.95 for identifying known lineage driver TFs across four hematopoietic lineages. Predictions validated by CRISPR-Cas9 knockout and single-cell Perturb-seq in zebrafish neural crest.

🔬 Applications and Insights

1️⃣ In Silico Perturbation Screening

Masking regulons and comparing velocity fields lets researchers simulate gene knockouts computationally, predicting cell fate shifts before running wet-lab experiments.

2️⃣ Cell Fate Driver Discovery

Applied to zebrafish neural crest, RegVelo identified tfec as a key early driver and elf1 as a regulator of pigment cell fate, both validated in vivo with CRISPR-Cas9.

3️⃣ Lineage-Specific GRN Recovery

In hematopoiesis, RegVelo recovered known lineage drivers (Smarca1, Pdx1, Mnx1, Hhex) across four lineages with high ranking accuracy (AUROC = 0.95).

4️⃣ Uncertainty-Aware Predictions

As a Bayesian generative model, RegVelo quantifies intrinsic and extrinsic cell state uncertainty, giving confidence estimates for both velocities and inferred regulatory edges.

💡 Why This Is Cool

This is the first framework to couple RNA velocity with gene regulatory networks in a single generative model. Rather than inferring dynamics and regulation separately and hoping they align, RegVelo learns them jointly. The result is a model that can simulate what happens when you perturb the regulatory wiring, with predictions validated from in silico all the way to in vivo knockouts. That closes the loop from computational hypothesis to experimental confirmation.

📄 Read the paper

💻 Try the code

TTM: Triplet Tumbling Microscopy Enables In Situ Quantification of Protein Complex Assembly and Dynamics

🔬 Protein-protein interactions drive nearly every cellular process, but measuring them inside living cells remains limited. FRET requires two labels and prior knowledge of interacting partners, while fluorescence anisotropy only works for small proteins below 50 kDa. There is no broadly applicable way to quantify protein complex size and binding dynamics in situ in real time.

TTM (Triplet Tumbling Microscopy) from Calico Life Sciences solves this by measuring rotational diffusion of protein complexes using only a single fluorescent tag. By leveraging long-lived triplet states in fluorescent proteins, TTM extends the measurable timescale from nanoseconds to hundreds of microseconds, covering the full range of cellular protein complexes.

🧬 TTM uses a pulsed excitation sequence: a 488 nm pulse generates triplet states aligned with the excitation polarisation, then an infrared trigger pulse (785-940 nm) reads out their orientation after a variable delay. As proteins tumble, the triplets lose alignment at a rate proportional to complex size. Rigid fluorescent protein tags (truncated mVenus and mStayGold) ensure tag motion faithfully reports target motion.

⚡ In purified protein systems, tumbling time constants scale linearly with molecular weight (r squared = 0.99). In living U2OS cells, TTM resolves complexes from 41 to 195 kDa from single-cell recordings (r squared = 0.85). It detects the approximately 10% size change from E6AP binding HPV16 E6 protein, quantifies p53 homo-oligomerisation states across nine point mutations, and tracks rapamycin-induced dimerisation dynamics in real time at approximately 3 Hz.

🔬 Applications and Insights

1️⃣ Single-Tag Interaction Detection

Unlike FRET, TTM requires only one fluorescent label and no prior knowledge of binding partners, making it applicable to uncharacterised or unexpected interactions.

2️⃣ Live-Cell Binding Dynamics

Real-time imaging at approximately 3 Hz captures the kinetics of complex formation as it happens, not just endpoint measurements.

3️⃣ Oligomerisation State Profiling

TTM distinguishes monomers, dimers, and tetramers of p53 in cells, revealing how tetramerisation domain mutations shift the equilibrium between functional states.

4️⃣ Standard Microscope Compatibility

The hardware requirements (pulsed lasers and an intensified camera) are compatible with most fluorescence microscopes, lowering the barrier to adoption across labs.

💡 Why This Is Cool

TTM fills a gap that has persisted for decades in cell biology: measuring how big a protein complex is inside a living cell, in real time, with a single label. The ability to track binding dynamics and oligomerisation states at physiological concentrations opens the door to studying protein interactions in their native context rather than in lysates or reconstituted systems. One tag, one measurement, real answers.

📄 Read the paper

Path2Space: AI-Predicted Spatial Transcriptomics Unlocks Breast Cancer Biomarkers from Pathology

🔬 Spatial transcriptomics is transforming our understanding of tumour heterogeneity, but its high cost limits large-scale biomarker discovery. Previous efforts to predict gene expression from histopathology slides have been restricted to small gene sets, precluding survival and treatment response analyses in large clinical cohorts.

Path2Space from NIH’s National Cancer Institute and Cedars-Sinai predicts the spatial expression of thousands of genes directly from routine H&E-stained histopathology slides. Trained on extensive breast cancer spatial transcriptomics data, it outperforms 21 established methods and enables scalable biomarker discovery without molecular assays.

🧬 Path2Space uses CTransPath, a digital pathology foundation model, to extract features from colour-normalised tile images around each spatial transcriptomics spot. A multilayer perceptron predicts log-transformed expression for 14,068 genes per spot. A spatial smoothing step averages predictions with neighbouring spots to mitigate technical variability. Trained on the Bassiouni et al. cohort comprising 56,567 matched image-expression spot pairs from 14 patients.

⚡ Median gene-wise PCC of 0.38 (smoothed) across 14,068 genes, with 6,629 genes exceeding PCC > 0.4. Binary classification of high versus low expression yields median AUC of 0.70, with 3,116 genes surpassing 0.75. Generalises robustly across three independent external cohorts (HEST, Martinez, HTAN). Applied to 976 TCGA breast cancer patients, Path2Space identifies three prognostic SpatioTypes and predicts chemotherapy and trastuzumab response at accuracy levels equal to or exceeding bulk sequencing biomarkers.

🔬 Applications and Insights

1️⃣ Low-Cost Spatial Biomarker Discovery

Derives spatial gene expression landscapes from routine pathology slides without expensive molecular assays, enabling large-cohort studies previously limited by cost.

2️⃣ Prognostic Breast Cancer Subtyping

Unsupervised clustering of predicted spatial transcriptomic profiles identifies three SpatioTypes with distinct biology and survival outcomes across 976 patients.

3️⃣ Treatment Response Prediction

Spatial biomarkers from H&E slides predict response to chemotherapy and trastuzumab at accuracy levels matching or exceeding those from bulk tumour sequencing.

4️⃣ Archival Tissue Applicability

Works on standard FFPE and fresh-frozen archival tissue, meaning existing hospital slide collections can be retrospectively analysed without new sample collection.

💡 Why This Is Cool

Spatial transcriptomics has been too expensive to run on the thousands of patients needed for robust biomarker discovery. Path2Space sidesteps this entirely by inferring spatial gene expression from H&E slides that hospitals already collect for every tumour. Turning routine pathology into a spatial omics readout could democratise precision oncology for any institution with a slide scanner.

📄 Read the paper

💻 Try the code

🗓️ Events & Competitions

The best competitions, hackathons, and community challenges in AI x life sciences, curated weekly. Know something worth featuring? Reply and let us know.

More upcoming events:

London Protein Design Day | June 23, Imperial College London

The first edition of a one-day symposium bringing together London’s protein design community and beyond. Programme spans AI-driven design, molecular dynamics, and bioinformatics, with applications across enzymes, antibodies, and materials. Organised by Pietro Sormanni, Rebecca Birolo, and Jakub Lála. Abstract deadline for poster/oral presentations is this Saturday (May 17). In person only.

BioHackathon Europe 2026 | November 9-13, Barcelona

ELIXIR’s annual international bioinformatics hackathon, running since 2018. 160+ participants, five days of collaborative coding on open bioinformatics infrastructure and tools. The call for project proposals opens March 16 and closes April 15 - so if you want to lead a project, that’s your window.

Thanks for reading!

💬 Get involved

We’re always looking to grow our community. If you’d like to get involved, contribute ideas or share something you’re building, fill out this form or reach out to me directly.

Connect With Us

Have questions or suggestions? We'd love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

Keeping up with AI x life science news can get exhausting.

It’s scattered across LinkedIn, X, Substack, arXiv, Slack, newsletters... and you still somehow miss the things that actually matter. Too much noise, not enough signal.

We’re building something to fix that: a smarter, more powerful way to stay on top of what’s actually relevant to you.

But we want to build it with you, not just for you. Take 2 minutes to tell us what’s missing. What you share will directly shape what we build, and you’ll be the first to benefit from it.

Share your input

Today we’re looking at Adaptyv, a Lausanne-based startup that’s building the experimental infrastructure the protein design revolution has been missing. We spoke with Tudor, who leads community and protein engineering efforts at Adaptyv, about why designing proteins computationally was only ever half the problem, and what it takes to close the gap between prediction and proof.

“People are realising now that data is the bottleneck for what we can currently achieve in ML for protein design. They need specialised functional data and they need to generate it fast.”

— Tudor, Adaptyv

🔴 The Problem

AI protein design has exploded. Diffusion models, language models, and structure prediction tools can now generate novel protein sequences in minutes. But there’s a persistent bottleneck that sits downstream of all that computation: actually testing whether the designs work.

Before Adaptyv, a computational protein designer who wanted to validate their binders had limited options. You could work with a contract research organisation, with complex onboarding, long timelines, and months before you saw results. You could do it yourself if you were lucky enough to have lab access. Or you could be in one of a handful of major labs, like the Baker lab, that had the infrastructure to run binding affinity measurements at scale.

For everyone else, the growing wave of independent protein designers, small academic groups, and early-stage biotechs training their own generative models, experimental validation was a wall. You could design as many proteins as you wanted on a computer, but you had no efficient way to know which ones actually folded, bound their target, or did anything useful.

“It was quite a black and white situation,” Tudor explains. “You were either in one of the bigger labs or you weren’t. And if you weren’t, you were designing stuff on the computer with no idea what it does in the real world.”

💡 The Idea

Adaptyv’s answer: a cloud lab purpose-built for protein designers.

The thesis has been core to the company since its founding by Julian Englert and Daniel Nakhaee-Zadeh, both engineers from EPFL. They initially built a microfluidics platform for high-throughput antibody-antigen binding testing, but when that approach proved too experimental, they pivoted. First to BLI and SPR-based measurement approaches in late 2023, then to an official platform launch in 2024.

The model is simple. You sign up on the Foundry platform. Upload your sequences via CSV. Add details: tags, antibody formats, terminal modifications. Select your target from a catalogue sourced from multiple suppliers. Submit. Adaptyv handles everything else.

The Foundry experiment creation workflow: choose your desired assay, choose targets for binding experiments, number of replicates, upload your sequences, and get an experiment quote for your draft experiment.

Behind the interface sits what the team calls LabOS, an orchestration brain that manages scheduling, expression, measurement, curve fitting, quality control, and results delivery across their automated systems. Users get back visualisations, raw sensorgrams, fitted binding curves, and all underlying data in a single package.

The key technical innovation is miniaturisation. By running cell-free expression reactions in microlitre volumes rather than large batch cultures, Adaptyv achieves roughly a 1,000x reduction in reagent use. No recombinant E. coli. No multi-day protein expression waits. Just small, fast, cell-free reactions that produce enough protein to measure on BLI or SPR.

📊 The Data

Current turnaround is approximately two weeks: one week for target QC if the target is new, then one week for the experiments themselves. The goal is to push that down to two to three days, and ultimately to near-instant validation that matches the speed of ML training cycles.

“Ideally, you want to get the experimental validation within the same time pressure as a training run,” Tudor says. “Almost instant. If we could make it faster, we would.”

The platform isn’t just a validation service; it’s becoming a data engine. Some customers are already running active learning loops: generate a set of designs, send them to Adaptyv, get results, retrain, repeat. Others are running large-scale campaigns to map the druggability of entire target spaces.

A typical use case: a team trains a diffusion model on PDB data, generates novel binders, and validates them through Adaptyv. The experimental results feed back into the next round of model training. Each cycle produces better designs and richer data.

“Both work hand in hand,” Tudor explains. “Now people are using us primarily for validation. But ultimately we want to go in the direction of large-scale custom data generation campaigns. We want to be a protein data centre.”

🔬 Proteinbase

The data Adaptyv was generating led naturally to their second platform: Proteinbase.

The problem it addresses is fragmentation. Existing protein databases use different protocols, different measurement methods, and different standards. Data from one source is hard to compare with data from another. Teams trying to train models on aggregated functional protein data spend enormous effort just standardising datasets, and even then, results often don’t reproduce.

“People were saying that what they’re missing is a unified database for functional proteins with standardised protocols,” Tudor says. “Everyone was trying to standardise datasets from different sources and running into the same problems. At least if everything is from a single source, it’s better for training better models.”

Proteinbase hosts competition data, provides unified downloads, and is building out community features: leaderboards, badges, knowledge sharing. It’s part database, part competition platform, part community hub.

The competitions have been a major growth driver. Recent rounds attracted over 600 participants, with hit rates up to 13% and binders in the nanomolar range. Adaptyv has also been hosting hackathons internationally in San Francisco and Berlin, with more planned across Europe. A recent GEM Bio Workshop competition hosted at ICLR on a disordered target (RBX1) drew more than 180 submissions.

💊 Who It’s For

Adaptyv serves three overlapping audiences: academic groups training and validating protein design models, biotech and pharma teams running lead optimisation or target characterisation campaigns, and independent protein designers who previously had no access to experimental validation.

Pricing is transparent and visible on the platform.

The philosophy is democratisation. Protein design has historically been concentrated in a small number of elite labs. Adaptyv is opening that up, not just through the lab infrastructure but through the community.

“Protein design has been super insular,” Tudor says. “It was either you were in one of the bigger labs or you weren’t. Giving people access to experimental validation and a community to share what they’ve learned, that’s how you grow the field.”

🔮 The Future

The next 12 months are focused on speed and automation.

On the Adaptyv platform: two-day experimental results, an expanded assay array, more standardisation, and deeper investment in agent-based workflows. The API is already designed for programmatic experiment submission, and the team is building towards a future where AI agents can control every aspect of the pipeline, from sample handling to liquid handlers to results retrieval.

“We have a two-level system,” Tudor explains. “A higher-level customer API for submitting experiments and getting results, and a lower-level API for every single machine interaction. In the future, agents could control all of it.”

On Proteinbase: more visible content, international hackathon expansion across Europe, and continued growth of the competition platform. The demand is clear: organisers are reaching out to Adaptyv to add protein design tracks to their events, and recent competitions have been oversubscribed.

The longer-term vision ties both platforms together. As the field recognises that data, not compute or model architecture, is the binding constraint on progress in ML for protein design, Adaptyv is positioning itself as the infrastructure layer that generates, validates, and distributes that data at scale.

Design is only half the problem. Adaptyv is building the other half.

👨‍🔬 Get in touch with Tudor.

💻 Adaptyv Website.

🌐 Adaptyv on LinkedIn.

🧬 Explore Proteinbase.

Thanks for reading Kiin Bio Weekly!

💬 Get involved

We’re always looking to grow our community. If you’d like to get involved, contribute ideas or share something you’re building, fill out this form or reach out to me directly.

Share Kiin Bio Weekly

Subscribe now to stay at the forefront of AI in Life Science and keep up with this upcoming season of deep dives.

Connect With Us

Have questions on this or suggestions for our next deep dive? We’d love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

Keeping up with AI x life science news can get exhausting.

It’s scattered across LinkedIn, X, Substack, arXiv, Slack, newsletters... and you still somehow miss the things that actually matter. Too much noise, not enough signal.

We’re building something to fix that: a smarter, more powerful way to stay on top of what’s actually relevant to you.

But we want to build it with you, not just for you. Take 2 minutes to tell us what’s missing. What you share will directly shape what we build, and you’ll be the first to benefit from it.

Share your input

🇺🇸 We’re heading to Bio-IT World in Boston, May 19-21.

Our CEO Filippo and CTO Bogdan will be there and would love to meet anyone thinking about:

• How AI is actually changing preclinical workflows (not just the hype)

• Why drug discovery is a systems problem, not just a science one

• What it takes to go from 5-year timelines to something radically faster

No pitch, just good conversation. If any of that’s on your mind, reach out - we’ll find a time to grab a coffee.

MolJSON: Molecular Representations for Large Language Models

🔬 LLMs are increasingly used in chemistry for tasks like reaction prediction and structure elucidation, but they need to read and write molecular structures reliably. Previous work has defaulted to SMILES strings or IUPAC names, but no one has systematically tested whether these formats are actually good for LLMs. Both impose strict serialisation rules that may not align with how language models process information.

Researchers at Oxford introduce MolJSON, a structured JSON schema that represents molecular graphs explicitly as lists of atoms and bonds. Unlike SMILES (which requires a specific graph traversal) or IUPAC (which requires rule-based nomenclature), MolJSON presents the molecular graph directly in a format compatible with LLM structured output modes.

🧬 They evaluated five molecular representations across 78,045 algorithmically generated questions spanning translation, shortest-path reasoning, and constrained generation tasks using GPT-5-nano, GPT-5-mini, GPT-5, and Claude Haiku 4.5. MolJSON consistently outperformed all existing formats as both an input and output representation.

⚡ On translation tasks, GPT-5 achieved 71.0% accuracy converting IUPAC to MolJSON versus 43.7% for IUPAC to SMILES. For constrained generation, GPT-5 reached 95.3% accuracy generating MolJSON versus 64.0% for SMILES and 76.3% for IUPAC. MolJSON was also 1.8x more token-efficient than SMILES on reasoning tasks. Performance advantages held even though models were never explicitly trained on MolJSON.

🔬 Applications and Insights

1️⃣ Better Chemistry Agents

LLM-based chemistry systems that read and write molecules can operate more reliably by switching to MolJSON, reducing errors from format parsing failures.

2️⃣ Robust Molecular Reasoning

MolJSON maintained high accuracy even on complex molecules with fused rings and high atom counts, where SMILES and IUPAC performance degraded sharply.

3️⃣ Token-Efficient Representations

Graph-based formats let models skip the internal reconstruction step needed for traversal-based representations, using fewer reasoning tokens and reducing latency.

4️⃣ Format-Agnostic Improvement

MolJSON outperformed SMILES and IUPAC despite those formats being well-represented in LLM training data, suggesting explicit graph encodings are intrinsically better aligned with how LLMs process structure.

💡 Why This Is Cool

Everyone building LLM chemistry tools has been using SMILES because it was already there. This paper shows that is leaving significant performance on the table. A simple change in molecular representation, with no model retraining, unlocks dramatically better accuracy across translation, reasoning, and generation. The fact that LLMs spontaneously generate semantically meaningful atom identifiers (like “C_acyl” or “Npip”) in MolJSON suggests these models can reason about molecular graphs more naturally when given an explicit graph format.

📄 Read the paper

💻 Try the code

PLaTITO: Protein Language Model Embeddings Improve Generalisation of Implicit Transfer Operators

🔬 Molecular dynamics simulations are essential for understanding protein behaviour, but conventional MD is computationally prohibitive for biologically relevant timescales. Generative molecular dynamics methods learn surrogate models from trajectory data, but they typically require large collections of long MD trajectories and struggle to generalise to unseen protein systems.

PLaTITO from Chalmers, Copenhagen, and DTU introduces coarse-grained transferable implicit transfer operators (TITO) for protein molecular dynamics that generalise to out-of-distribution protein systems. By conditioning on protein language model embeddings from ESM and structure embeddings from Proteina, the model learns to transfer across diverse proteins without system-specific fine-tuning.

🧬 Trained on the mdCATH dataset (4,471 domains, ~56 ms aggregate simulation time), PLaTITO learns long-time transition densities conditioned on backbone coordinates, sequence, temperature, and time step. The architecture uses a two-stage design: a conditioning network produces per-residue representations, and a velocity network generates the flow field for sampling future conformations.

⚡ PLaTITO-Big (19M parameters) outperforms BioEmu across all equilibrium sampling metrics on fast-folding proteins while requiring substantially less training data (56 ms vs. 216 ms) and compute (1,100 GPU hours vs. 9,216). It recovers non-Arrhenius temperature-dependent folding kinetics and explores cryptic binding pockets, generating trajectories with repeated folding and unfolding events at microsecond timescales.

🔬 Applications and Insights

1️⃣ Out-of-Distribution Generalisation PLaTITO transfers to unseen proteins without fine-tuning, unlike Boltzmann Emulators that require system-specific training data.

2️⃣ Data-Efficient Training Achieving state-of-the-art equilibrium sampling with 4x less MD data and 8x less compute than BioEmu demonstrates that transfer learning can dramatically reduce the data barrier.

3️⃣ Temperature-Dependent Kinetics Explicitly conditioning on temperature lets PLaTITO capture non-Arrhenius behaviour, reproducing physically meaningful folding and unfolding rates across temperature ranges.

4️⃣ Cryptic Binding Pocket Discovery PLaTITO-Big samples conformational transitions to cryptic pockets from both apo and holo states, opening applications in drug discovery for targeting hidden binding sites.

💡 Why This Is Cool

This is the first transferable molecular dynamics model that genuinely generalises across protein systems while beating dedicated Boltzmann Emulators on their own benchmarks. The key insight is that pretrained protein language models encode enough structural and evolutionary information to let a small (3-19M parameter) dynamics model transfer across diverse folds. Generating realistic microsecond-scale folding trajectories on a single GPU in seconds, rather than months of conventional simulation, changes what is computationally accessible for studying protein dynamics.

📄 Read the paper

OpenBind: A Structure-Affinity Dataset for Structure-Based AI in Drug Discovery

🔬 Structure-based AI for drug discovery is held back by a data bottleneck. Public protein-ligand datasets are sparse, unevenly distributed, and rarely link crystallographic binding modes with quantitative affinity measurements at scale. Current ML methods for docking, cofolding, and affinity prediction are difficult to evaluate fairly because most benchmarks overlap with training data.

The OpenBind consortium (Diamond Light Source, Oxford, and partners) releases a dense structure-affinity dataset: 925 crystallographic binding events from 699 compounds with affinity measurements for 601 compounds, all targeting EV-A71 2A protease, a viral target relevant to pandemic preparedness. The data are deliberately positioned in an under-represented region of protein-ligand space relative to existing public structures.

🧬 The dataset includes fragment screen hits and follow-on compounds with KD values from Creoptix WAVEsystem measurements, creating a coherent experimental series where users can study local structure-activity relationships. Reference benchmarks span conventional docking (AutoDock Vina), ML docking (GNINA, DiffDock), cofolding (AlphaFold3, Boltz, OpenFold3), and affinity prediction methods.

⚡ Redocking achieves up to 85% success (GNINA), but cross-docking into apo structures drops below 5% due to binding-site loop conformational changes. Cofolding methods reach 36% (OpenFold3-p2), but fine-tuning on fragment-screen data boosts this to 76%, approaching redocking performance. For affinity prediction, a simple molecular-weight baseline outperforms most structure-based methods, highlighting how challenging this task remains.

🔬 Applications and Insights

1️⃣ Benchmarking Beyond Training Data

The EV-A71 2A protease complexes are dissimilar to pre-2021 PDB data, providing a genuine test of whether cofolding and docking methods generalise or just memorise near-neighbours.

2️⃣ Fragment Screens as Training Data

Fine-tuning cofolding models on fragment-bound structures doubled success rates on follow-on compounds, showing that early experimental data can feed directly into AI model improvement.

3️⃣ Separating Failure Modes

The dataset cleanly distinguishes receptor-conformation failures from ligand-placement failures, letting method developers target specific weaknesses rather than debugging aggregate metrics.

4️⃣ Affinity Prediction Reality

Check Simple baselines beating structure-based methods on this dataset is a clear signal that current affinity models may be learning chemical trends rather than genuine protein-ligand interaction physics.

💡 Why This Is Cool

OpenBind is not just releasing more structures. It is building the experimental infrastructure to generate the kind of data that structure-based AI actually needs: dense, linked structure-affinity measurements within coherent chemical series, positioned where current models are weakest. The fragment fine-tuning result is particularly striking. It shows that a relatively small crystallographic screen can transform cofolding performance on a new target, pointing toward a practical workflow where early experiments directly improve computational predictions for the same campaign.

📄 Read the data

💻 Try the benchmarks

📬 Newsletter Shout-Out

This week we're shouting out Building in BioAI, a monthly newsletter from Joe:

Building in BioAI is a monthly newsletter for those operating in, or interested in, the AI-enabled biology space. That’s founders, technical leaders, and individual contributors working within areas like therapeutics, diagnostics, and tooling.

Joe’s roundup centres on observations from within the space, including analysis of how teams are structuring themselves, what’s changing in hiring, where funding is landing, what headlines mean for growth, and how BioAI companies are thinking about commercialising what they’re building.

Each edition pulls from ongoing conversations with people doing the work day-to-day, as well as his own take on what’s hit headlines that month.

Joe recruits in this space day-to-day, and so often speaks from that vantage point. He spends most of his time inside these teams, hiring for them, speaking with founders and senior talent across the market. The aim isn’t to overstate where things are going, but to give a clear picture of what’s actually happening and why it matters if you’re hiring or looking to commercialise in BioAI.

🔗 Check it out!

🗓️ Events & Competitions

The best competitions, hackathons, and community challenges in AI x life sciences, curated weekly. Know something worth featuring? Reply and let us know.

More upcoming events:

BioHackathon Europe 2026 | November 9-13, Barcelona

ELIXIR’s annual international bioinformatics hackathon, running since 2018. 160+ participants, five days of collaborative coding on open bioinformatics infrastructure and tools. The call for project proposals opens March 16 and closes April 15 - so if you want to lead a project, that’s your window.

Thanks for reading!

💬 Get involved

We’re always looking to grow our community. If you’d like to get involved, contribute ideas or share something you’re building, fill out this form or reach out to me directly.

Connect With Us

Have questions or suggestions? We'd love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

Keeping up with AI x life science news can get exhausting.

It’s scattered across LinkedIn, X, Substack, arXiv, Slack, newsletters... and you still somehow miss the things that actually matter. Too much noise, not enough signal.

We’re building something to fix that: a smarter, more powerful way to stay on top of what’s actually relevant to you.

But we want to build it with you, not just for you. Take 2 minutes to tell us what’s missing. What you share will directly shape what we build, and you’ll be the first to benefit from it.

Share your input

Around 30% of clinical trials fail due to safety concerns. That’s billions in sunk costs, years of lost time, and patients who don’t get the medicines they need. Sable Bio thinks the problem starts much earlier in the pipeline, with how safety assessment is done in the first place.

I spoke with Olly Oechsle, CTO of Sable Bio, about how time-consuming traditional safety workflows can be and how the company’s Target Intelligence platform helps toxicologists.

🔴 The Problem

Before a drug candidate moves into preclinical development, safety scientists need to answer a deceptively simple question: if we inhibit or activate this target, what else is going to happen that we didn’t intend?

Answering that means looking across clinical trial data, genetic association studies, mouse knockout phenotypes, scientific literature, expression profiles, and more. Each lives in a different database, requires different expertise to interpret, and few are built with a toxicologist’s specific needs in mind.

The result is a process that takes anywhere from a few days to a month. Safety scientists spend a lot of that time collating information rather than doing the analytical work they’re trained for: identifying risks, building mitigation strategies, and making judgment calls about whether a target’s safety profile balances its therapeutic potential.

There’s also a reproducibility problem. “If another scientist were to do the same research with the same amount of time, would they have come up with the same answer?” Olly asks. When you’re manually searching PubMed, reviewing mouse knockouts or sifting through clinical databases, you inevitably go deep on some rabbit holes while missing others entirely. There’s no systematic way to know whether every potential adverse event has been examined across every relevant data source.

And the challenge doesn’t end with a single assessment. Drug discovery programs run for years. New data emerges, new papers are published, and safety scientists, often stretched across multiple projects, need to stay on top of all of it.

💡 The Platform

Sable’s core product is Target Safety Reports. A user searches for any target in the genome, specifies whether they plan to inhibit or activate it, and receives a comprehensive, customized safety report.

Findings can be organized by organ system (cardiovascular risk, hepatic risk, and so on) or by data source, letting users drill into the literature, clinical data, genetic associations, or expression profiles separately. The platform pulls from PubMed, clinical trial databases, GWAS data, gene burden studies, and MGI mouse knockout data, among other sources.

What makes this more than a search engine is how Sable has tuned each data source specifically for safety science and collated the result into a coherent report. For clinical trial data, statistical methods distinguish drug-centric effects from target-driven ones, and on-target from off-target effects, accounting for patient comorbidities that can be mistaken for causative drug effects. For the literature, Sable has built proprietary language models that extract target-to-adverse-event relationships with precision and coverage that general-purpose LLMs can’t match.

“It’s easy to make early strides with literature but it’s really hard to do it well,” Olly says. General AI models can answer questions about a single paper, but they struggle with corpus-wide analysis, and tend to “enthusiastically offer insights” from their broader training data that goes beyond the evidence being reviewed. Meanwhile literature tools can quickly present an overwhelming volume of content.

The platform also includes tracking that alerts users when new information emerges affecting a target’s safety profile, plus collaboration features for commenting, discussing, and sharing reports across teams.

📊 Weight of Evidence

Central to Sable’s approach is the toxicology principle of “weight of evidence.” Rarely does a single data source give a definitive answer on target safety. Instead, signals come from multiple directions: a suggestive mouse knockout phenotype, a mechanism described in the literature, expression data showing the target is active in a particular tissue, and adverse events observed with a related ligand.

Sable brings all of these signals together, letting scientists evaluate the full picture rather than chasing individual threads across separate databases. This works both ways: sometimes the different evidence types show a perceived hazard isn’t actually a concern, potentially saving organizations from expensive and unnecessary preclinical studies.

🧩 Where It Fits

Sable’s sweet spot is late discovery through to lead optimization, where the question shifts from “will this work?” to “what else is this doing?” The platform is used by preclinical safety scientists at several top-10 pharma companies, biotechs, venture capital firms conducting asset due diligence, and by consultants running target safety assessments.

The team is exploring expansion into earlier target selection, and for organizations using AI-driven target ID platforms that generate dozens or hundreds of candidates, Sable is developing wider-scale safety assessment products, systematically evaluating 50 to 200 targets with a traceable decision-making record. These are planned for release within the year.

Beyond the web interface, Sable offers API access and is building MCP integrations, positioning itself as a universal safety layer that plugs into the broader drug discovery ecosystem.

🔮 The Future

Sable is rolling out its new proprietary literature models alongside deeper analysis for single-cell data, expression data, and mechanistic biology. The company is also running a side-by-side comparison study between toxicologist assessments and platform outputs, with results expected in a couple of months.

“We’re not there to replace the decision-making process, which is ultimately a human one,” Olly says. “We’re there to save experts’ time.” Given how much of that time currently goes toward collecting data rather than acting on it, that’s a proposition most safety scientists can get behind.

🧑‍🔬Get in touch with Olly.

💻Sable Bio Website.

🌐Sable Bio on LinkedIn.

Thanks for reading Kiin Bio Weekly!

💬 Get involved

We’re always looking to grow our community. If you’d like to get involved, contribute ideas or share something you’re building, fill out this form or reach out to me directly.

Share Kiin Bio Weekly

Subscribe now to stay at the forefront of AI in Life Science and keep up with this upcoming season of deep dives.

Connect With Us

Have questions on this or suggestions for our next deep dive? We’d love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

Keeping up with AI x life science news can get exhausting.

It’s scattered across LinkedIn, X, Substack, arXiv, Slack, newsletters... and you still somehow miss the things that actually matter. Too much noise, not enough signal.

We’re building something to fix that: a smarter, more powerful way to stay on top of what’s actually relevant to you.

But we want to build it with you, not just for you. Take 2 minutes to tell us what’s missing. What you share will directly shape what we build, and you’ll be the first to benefit from it.

Share your input

🇺🇸 We’re heading to Bio-IT World in Boston, May 19-21.

Our CEO Filippo and CTO Bogdan will be there and would love to meet anyone thinking about:

• How AI is actually changing preclinical workflows (not just the hype)

• Why drug discovery is a systems problem, not just a science one

• What it takes to go from 5-year timelines to something radically faster

No pitch, just good conversation. If any of that’s on your mind, reach out - we’ll find a time to grab a coffee.

dEVA: Zero-Shot Design of a De Novo Metalloenzyme

🔬 Designing functional enzymes from scratch remains one of the hardest challenges in protein design. Previous approaches relied on borrowing catalytic motifs from nature, but optimising for structure alone does not guarantee catalytic competence. Efficient catalysis requires a precise balance of chemical, geometric, and electrostatic criteria that existing methods struggle to jointly satisfy.

Gina El Nesr and colleagues at Stanford present dEVA (design by EVolutionary Algorithm), a multi-objective framework built on NSGA-II that simultaneously optimises multiple design objectives, enriching for candidates where all criteria are mutually compatible rather than traded off against one another. Using LigandMPNN for sequence design and Metal3D for metal site prediction, dEVA iterates mutations across generations, converging on Pareto-optimal solutions.

🧬 Their best design, desB, achieves catalytic efficiency of 1,500 M⁻¹s⁻¹ and rate enhancement of 3x10¹³ without directed evolution. Analysis of PDB zinc sites revealed around 10% had zero coordinating ligands and over 54% had two or fewer residues, many being crystallisation artefacts. They trained Metal3D-Clean and Metal3D-Cat on curated data to address this.

⚡ desB hydrolyses both phosphomonoesters (uncatalysed half-life >500,000 years) and phosphodiesters (>13 million years). This promiscuity mirrors early enzyme evolution, opening the door to engineering specificity from a designed starting point.

🔬 Applications and Insights

1️⃣ Zero-Shot Functional Enzyme

Design dEVA designed a catalytically active metalloenzyme without natural templates or evolutionary information. No directed evolution was needed to achieve function.

2️⃣ Multi-Objective Optimisation Over Single-Score Ranking

By treating design as population-based evolutionary search across multiple objectives, dEVA avoids the compromises of single-score optimisation or sequential filtering.

3️⃣ Training Data Quality Matters

Curating PDB metal sites and retraining Metal3D on catalytically relevant examples was essential. Garbage in, garbage out applies to structural prediction too.

4️⃣ A Platform for Designed Catalysis

The promiscuous activity of desB mirrors early enzyme evolution, providing a starting point from which specificity can be engineered.

💡 Why This Is Cool

This is the first de novo enzyme designed without borrowing from nature that matches natural catalytic efficiency. The rate enhancement of 3x10¹³ is the highest for any de novo designed hydrolase. dEVA shows functional catalytic sites can emerge computationally from first principles.

📄 Read the paper

💻 Try the code

Expression Rescue: Structure-Guided Recovery of Antibody Productivity

🔬 High-affinity antibody variants often fail in production because of poor cellular expression. Researchers at Seoul National University combined AlphaFold3 and ProteinMPNN into a rescue workflow that identifies sequence-structure mismatches in CDR residues and corrects them, often with a single substitution, restoring expression while preserving binding affinity.

🧬 Affinity and expression are largely independent, meaning high-affinity, low-productivity (HALP) clones are not failures. They can be rescued.

⚡ ProteinMPNN scores correlate strongly with cellular productivity across around 9,500 variants, revealing that expression is governed by how well CDR sequences fit their structural context, not just their biochemical properties.

🔬 Applications and Insights

1️⃣ Rescuing High-Affinity Failures

Across 14 diverse HALP antibodies, single-residue substitutions restored expression in 11 cases, while preserving 80% or more of original binding affinity. This reframes failed candidates as recoverable assets rather than endpoints.

2️⃣ Sequence-Structure Compatibility as a Predictor

ProteinMPNN scores serve as a reliable proxy for expression, providing a computationally cheap filter before expensive experimental validation.

3️⃣ Minimal Edits, Maximal Impact

Rescue often required only one mutation, yielding up to 4-fold improvements in expression. These substitutions typically stabilise interactions within CDRs or between CDRs and the antibody framework.

4️⃣ Decoupling Affinity and Developability

Because productivity and affinity landscapes are independent, expression can be improved without compromising binding, solving a long-standing trade-off in antibody engineering.

💡 Why This Is Cool

Antibody design has long focused on finding better binders, but binding is only half the story. By reframing expression as a structural compatibility problem, failed candidates become fixable rather than disposable. This turns antibody engineering from a filtering step into a repair-and-optimise loop, expanding the usable therapeutic space.

📄 Read the paper

💻 Try the code

SyntheMol-RL: Reinforcement Learning for Designing Easily Synthesizable Antibiotics

🔬 Generative AI can propose drug candidates, but most fail at the same hurdle: they cannot be synthesised efficiently. Molecules that look promising on screen often require impractical chemistry to produce. SyntheMol-RL from Stanford and McMaster University uses reinforcement learning to navigate a chemical space of 46 billion synthesisable compounds, optimising for both antibacterial activity and aqueous solubility simultaneously.

🧬 Built on real chemical building blocks and validated reaction templates, the model generates molecules with guaranteed synthetic routes. This is not theoretical synthesisability. Every output comes with a concrete pathway from purchasable reagents.

⚡ 79 novel compounds were synthesised and tested. 13 showed potent in vitro activity against Staphylococcus aureus. Seven passed structural novelty filters against known antibiotics. One compound, synthecin, demonstrated efficacy in a murine wound infection model of methicillin-resistant S. aureus (MRSA).

🔬 Applications and Insights

1️⃣ From Screen to Syringe

Unlike virtual screening, SyntheMol-RL generates molecules with built-in synthesis plans, removing the bottleneck between computational hits and experimental validation.

2️⃣ Multi-Parameter Optimisation

Reinforcement learning enables simultaneous tuning of activity, solubility, and synthesisability rather than optimising one property at the cost of others.

3️⃣ In Vivo Validation

Synthecin’s efficacy in a wound infection model moves AI-designed antibiotics beyond petri dish activity into preclinical relevance.

4️⃣ Generalisable Framework

The architecture is target-agnostic. Swap the reward function and the same framework applies across therapeutic domains, not just antibiotics.

💡 Why This Is Cool

Most generative models for drug discovery propose molecules that cannot be made, or can be made but do not work. SyntheMol-RL closes both gaps: it only proposes what chemistry can deliver, and it validated a compound through to animal models. Going from 46 billion possibilities to a single molecule treating MRSA-infected wounds in mice is the full loop from generative AI to preclinical candidate.

📄 Read the paper

💻 Try the code

📬 Newsletter Shout-Out

This week we're shouting out Building in BioAI, a monthly newsletter from Joe:

Building in BioAI is a monthly newsletter for those operating in, or interested in, the AI-enabled biology space. That’s founders, technical leaders, and individual contributors working within areas like therapeutics, diagnostics, and tooling.

Joe’s roundup centres on observations from within the space, including analysis of how teams are structuring themselves, what’s changing in hiring, where funding is landing, what headlines mean for growth, and how BioAI companies are thinking about commercialising what they’re building.

Each edition pulls from ongoing conversations with people doing the work day-to-day, as well as his own take on what’s hit headlines that month.

Joe recruits in this space day-to-day, and so often speaks from that vantage point. He spends most of his time inside these teams, hiring for them, speaking with founders and senior talent across the market. The aim isn’t to overstate where things are going, but to give a clear picture of what’s actually happening and why it matters if you’re hiring or looking to commercialise in BioAI.

🔗 Check it out!

🗓️ Events & Competitions

The best competitions, hackathons, and community challenges in AI x life sciences, curated weekly. Know something worth featuring? Reply and let us know.

More upcoming events:

BioHackathon Europe 2026 | November 9-13, Barcelona

ELIXIR’s annual international bioinformatics hackathon, running since 2018. 160+ participants, five days of collaborative coding on open bioinformatics infrastructure and tools. The call for project proposals opens March 16 and closes April 15 - so if you want to lead a project, that’s your window.

Thanks for reading!

💬 Get involved

We’re always looking to grow our community. If you’d like to get involved, contribute ideas or share something you’re building, fill out this form or reach out to me directly.

Connect With Us

Have questions or suggestions? We'd love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

Keeping up with AI x life science news can get exhausting.

It’s scattered across LinkedIn, X, Substack, arXiv, Slack, newsletters... and you still somehow miss the things that actually matter. Too much noise, not enough signal.

We’re building something to fix that: a smarter, more powerful way to stay on top of what’s actually relevant to you.

But we want to build it with you, not just for you. Take 2 minutes to tell us what’s missing. What you share will directly shape what we build, and you’ll be the first to benefit from it.

Share your input

Today we’re looking at Rowan, a Boston-based startup of six that’s building a web-based computational chemistry platform that lets medicinal chemists run advanced modeling workflows directly, without needing to manage scripts, infrastructure, or specialist software.. We spoke with co-founder Corin Wagen, an experimental organic chemist turned computational entrepreneur, about why the gap between medicinal chemists and computational tools has persisted for so long, and what it takes to close it.

“There are all these really smart, really talented chemists and scientists who are just not able to use computation to help them out. You always have to ask somebody else to do it. There’s all these artificial barriers.”

— Corin Wagen, Co-founder, Rowan

🔴 The Problem

Computational chemistry has a usability problem.

The tools exist. You can predict binding affinities, generate conformers, run molecular dynamics, dock compounds into protein structures. But actually using any of this typically requires programming expertise, command-line fluency, access to the right hardware, and the patience to stitch together a dozen different software packages that were never designed to work together.

For most medicinal chemists, the people actually deciding which compounds to synthesise next, this means going through a computational chemist every time they want to run something. That handoff slows everything down. It creates bottlenecks, introduces miscommunication, and means that computation gets used selectively rather than routinely.

The result: most drug discovery teams are making synthesis decisions with less computational insight than they could be: not because the science isn’t there, but because the software gets in the way.

“You should be a chemist to use Rowan, but you shouldn’t need to be a programmer.”

💡 The Idea

Rowan’s answer is a web-based platform that organises computational chemistry into workflows, data in, data out, aligned to how scientists actually think about experiments.

You have a compound and want to know how soluble it will be? That’s a workflow. You have a binding pose and want to screen 100 analogues? That’s a workflow. Under the hood it might be an ML model, a physics simulation, or a database lookup, but the scientist doesn’t need to care about the plumbing.

The founding team brings a deliberate mix: machine learning, software engineering (ex-Meta), quantum chemistry, product and business, and experimental organic chemistry. Wagen sees that breadth as essential. “A lot of times the best ideas in chemistry come from people who’ve journeyed outside chemistry and then bring back new ideas.”

The platform spans workflows from hit discovery to candidate selection, ligand-based methods (ML potentials, rapid quantum chemistry, conformers, reactivity, spectra) and, increasingly, structure-based drug discovery (docking, co-folding, molecular dynamics, and now free energy perturbation). Scientists pick what they need. Rowan doesn’t prescribe a single workflow.

Critically, Rowan is designed to fit into existing scientific software stacks. Teams can use it through the browser, through Python, or as part of agentic and automated pipelines, without replacing the tools they already rely on.

“There’s a lot more smart people outside Rowan than inside Rowan. We don’t need to own the whole thing.”

⚙️ The FEP Release

The headline addition is free energy perturbation (FEP), the gold-standard physics-based method for predicting how binding affinity changes across a series of related compounds. If you’re optimising a lead and need to decide which of 100 analogues to actually synthesise, FEP tells you which ones are likely to bind better and which are duds, before you spend time and money making them.

FEP has been around for decades, but two things have kept it out of mainstream medicinal chemistry workflows: it’s expensive (historically around 10 GPU hours per compound) and it’s complicated to set up and run.

Rowan partnered with Forrest York and the open-source TMD engine (originally from Relay Therapeutics) to tackle both. The engineering improvements are dramatic: default settings run approximately 10 minutes per leg, with the potential to reach 1-2 minutes per leg with adjusted settings. That’s a roughly 60x speedup over the literature standard.

The speed comes from low-level optimisation and an algorithmic trick called local resampling, where instead of simulating the entire protein, the calculation focuses on the immediate neighbourhood around the ligand. “If you naively try to do that, it works very poorly,” Wagen explains. “But it turns out if you very cleverly try to do that, it works very well.”

The end-to-end workflow can be run without any coding. Three steps: prepare your poses using analogue docking, build a perturbation graph showing which ligands to compare, and run the FEP calculations on cloud GPUs. Results stream back to your browser in real time.

📊 The Trade-Off

The honest question: how accurate is it?

Rowan’s benchmarks show a mean absolute error of approximately 1.3 kcal/mol, compared to Schrödinger’s ~0.8 kcal/mol. The gap comes from two places: the speed-optimising approximations and the use of open force fields rather than Schrödinger’s proprietary ones.

Wagen is transparent about this. “Our benchmarks are a little bit worse than Schrödinger’s. They have amazing force fields. They’re so good at force fields.”

But the metric that matters most in practice isn’t absolute energy prediction, it’s ranking. If FEP correctly tells you which compounds will bind better and which won’t, you’ve saved synthesis cycles regardless of whether the exact energy numbers are perfect. And on ranking, the performance gap narrows considerably.

“If compounds are predicted to bind terribly, unless you really don’t understand the binding mode, they’re almost always bad binders. We have a lot of benchmark data showing this.”

The calculus is straightforward: if you can run FEP on 10–100x more compounds than you can synthesise, at a fraction of the cost and time, slightly wider error bars are a trade-off most teams will take.

Rowan publishes its benchmark data openly: you can explore the full results at benchmarks.rowansci.com.

🔬 Why It’s Different

Speed changes behaviour. At 10 GPU hours per compound, FEP is something you run occasionally on high-value decisions. At 10 minutes, it becomes routine, something you run on every idea before committing to synthesis. That shift from selective to systematic is the real unlock.

No code, no setup. Most FEP implementations require protein preparation scripts, force field configuration, graph construction code, and hardware management. Rowan handles all of this in three browser-based workflows, while also providing API access for teams that want to automate and run FEP at scale.

Open infrastructure, not a black box. Built on the open-source TMD engine and open force fields. Benchmarks are published. The platform is designed to complement existing tools via API, not replace entire workflows.

Built for the medicinal chemist. Rowan is built to put computational insight directly in the hands of medicinal chemists, while still remaining useful to computational and platform teams.

💊 Who It’s For

Rowan’s primary customers are small-to-medium biotechs and pharma departments that don’t have large internal computational tooling teams. The philosophy is to complement, not compete.

“We often work with companies where they’re small enough that they don’t have anybody who they can ask to do the work that we do for them,” Wagen says.

The platform recently passed 10,000 users and has generated over 40 publications. Pricing for FEP runs at approximately $5-10 per edge for platform customers, or $25 per ligand through a managed fee-for-service option where a Rowan scientist handles the analysis.

🔮 The Future

The near-term goal is integration into fast-moving drug discovery programmes through active pilots. The dream: automated nightly runs from SMILES strings to predicted binding affinities, with a digest landing in your inbox each morning showing which AI-generated analogues are worth pursuing.

“I suspect that the dream is that we can just blindly put SMILES in, go all the way to binding affinities, and that runs every single night.”

That dream is not yet reality. Fully automated structure-based modeling is still not trivial, particularly in pose preparation and graph construction, and Wagen is candid that the final stretch depends on solving messy, program-specific edge cases and learning from their early FEP customers. Beyond automation, the roadmap includes custom force-field fitting for client compound series (to close the accuracy gap),, pre-FEP triage tools, and continued speed optimisation targeting under one minute per compound for large libraries.

“I hope this doesn’t take us a few years. I hope this can happen in 2026 for Rowan.

👨‍🔬 Get in touch with Corin

💻 Rowan Website.

🌐 Rowan on LinkedIn.

📺 Watch the FEP walkthrough on YouTube.

📄 Read the FEP core concepts explainer.

Thanks for reading Kiin Bio Weekly!

💬 Get involved

We’re always looking to grow our community. If you’d like to get involved, contribute ideas or share something you’re building, fill out this form or reach out to me directly.

Share Kiin Bio Weekly

Subscribe now to stay at the forefront of AI in Life Science and keep up with this upcoming season of deep dives.

Connect With Us

Have questions on this or suggestions for our next deep dive? We’d love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

Keeping up with AI x life science news can get exhausting.

It’s scattered across LinkedIn, X, Substack, arXiv, Slack, newsletters... and you still somehow miss the things that actually matter. Too much noise, not enough signal.

We’re building something to fix that: a smarter, more powerful way to stay on top of what’s actually relevant to you.

But we want to build it with you, not just for you. Take 2 minutes to tell us what’s missing. What you share will directly shape what we build, and you’ll be the first to benefit from it.

Share your input

🇺🇸 We’re heading to Bio-IT World in Boston, May 19-21.

Our CEO Filippo and CTO Bogdan will be there and would love to meet anyone thinking about:

• How AI is actually changing preclinical workflows (not just the hype)

• Why drug discovery is a systems problem, not just a science one

• What it takes to go from 5-year timelines to something radically faster

No pitch, just good conversation. If any of that’s on your mind, reach out - we’ll find a time to grab a coffee.

HEIST: A Graph Foundation Model for Spatial Transcriptomics and Proteomics

🔬 Spatial transcriptomics captures gene expression within tissue architecture, but existing models either ignore spatial relationships or flatten each cell into a simple feature vector. They miss the interplay between a cell’s internal gene programmes and its tissue neighbourhood.

HEIST from Yale models tissues as hierarchical graphs. The upper level captures spatial relationships between cells, while each cell is represented by its own gene co-expression network. Cross-level message passing connects the two, letting internal regulation and spatial context inform each other.

🧬 Pretrained on 22.3 million cells from 124 tissues across 15 organs using spatially-aware contrastive learning and masked autoencoding. HEIST uses flexible gene vocabularies rather than fixed gene sets, so it generalises to unseen genes and even spatial proteomics without retraining.

⚡ Unsupervised analysis reveals spatially informed cell subpopulations missed by prior models. Downstream, HEIST achieves state-of-the-art performance in clinical outcome prediction, cell type annotation, and gene imputation across multiple spatial technologies.

🔬 Applications and Insights

1️⃣ Cross-Modal Generalisation

Transfers from transcriptomics to proteomics without retraining, making it applicable across spatial profiling technologies.

2️⃣ Tissue Microenvironment Discovery

The hierarchical design captures spatially defined subpopulations that flat models miss, enabling more nuanced tissue phenotyping.

3️⃣ Clinical Outcome Prediction

Patient-level embeddings from spatial data support tasks like treatment response and survival prediction.

4️⃣ Flexible Gene Vocabularies

By avoiding fixed gene sets, HEIST handles new panels and custom targets without architectural changes.

💡 Why This Is Cool

Most spatial models look at where cells are or what they express. HEIST does both through hierarchical graph modelling. Generalising to proteomics without retraining suggests these representations capture something fundamental about how cells organise within tissues.

📄 Read the paper.

💻 Try the code.

The Synthetic Epitope Atlas: High-Throughput Design and Validation of De Novo Antibody-Antigen Complexes

🔬 ML models for antibody design are held back by a data bottleneck: not enough structural training data linking designed antibodies to validated binding outcomes. Existing datasets are small, biased towards natural antibodies, and lack systematic off-target measurements.

A-Alpha Bio built SEPIA (Synthetic Epitope Atlas), pairing over 26 million on-target and off-target binding measurements with computationally designed VHH-antigen structures. Using their AlphaSeq yeast-based platform, they measured binding affinities and specificity across thousands of de novo synthetic epitope proteins designed to bind VHH nanobodies.

🧬 Each designed VHH-SEP pair comes with both structural predictions and experimental binding data, so models can learn what makes a designed complex actually bind versus what only looks good computationally.

⚡ Across thousands of variants, SEPIA validates strong specificity and provides the negative data most antibody datasets lack. Positive and negative measurements at this scale give ML models a clearer signal for learning specificity, not just affinity.

🔬 Applications and Insights

1️⃣ Training Data for Antibody ML

26 million measurements paired with designed structures create a purpose-built resource for next-generation antibody design models.

2️⃣ Specificity, Not Just Affinity

Systematic off-target measurements let models learn what not to bind, addressing a major blind spot in current datasets.

3️⃣ Closing the Design-Validation Loop

Linking computational designs directly to high-throughput experimental readouts enables rapid iteration on antibody engineering.

4️⃣ Nanobody-Focused Design

VHH nanobodies are increasingly important as therapeutics. A large-scale, VHH-specific dataset accelerates this growing field.

💡 Why This Is Cool

The gap between computational antibody design and experimental reality has always been the data. SEPIA fills it with 26 million purpose-built binding measurements, including both what works and what does not. Models trained on real specificity data at this scale can finally learn to design antibodies that are specific, not just tight binders.

📄 Read the paper.

Apollo: A Multimodal Temporal Foundation Model for Virtual Patient Representations

🔬 Modern hospitals generate vast multimodal data across labs, imaging, notes, medications, and procedures, but it sits in disconnected systems. No existing model integrates the full breadth and temporal depth of a clinical record into one unified representation.

Apollo from Harvard Medical School does exactly that. Trained on over 30 years of longitudinal records from Mass General Brigham, it unifies 28 modalities and 12 major specialties into a shared embedding space, building an “atlas of medical concepts.”

🧬 Apollo processes entire care journeys as sequences of structured and unstructured events, compressing them into virtual patient representations. Its vocabulary spans over 100,000 unique medical events alongside clinical images and free-text notes. Feature attribution confirms predictions align with clinically interpretable biomarkers.

⚡ Evaluated across 322 tasks on 1.4 million held-out patients: disease onset prediction up to five years ahead (95 tasks), disease progression (78), treatment response (59), adverse event risk (17), and hospital operations (12). Apollo also functions as a multimodal medical search engine across 61 retrieval tasks.

🔬 Applications and Insights

1️⃣ Five-Year Disease Forecasting

Predicting disease onset years in advance from the full patient record enables proactive intervention rather than reactive care.

2️⃣ Treatment Response Prediction

Drawing on a patient’s complete multimodal history supports more personalised therapy decisions across specialties.

3️⃣ Multimodal Medical Search

Text and image queries against patient embeddings create a clinical search engine for cohort identification and case matching.

4️⃣ Interpretable Predictions

Feature attribution shows outputs align with known biomarkers, bridging AI predictions and clinical reasoning.

💡 Why This Is Cool

This is the first model to compress decades of multimodal clinical data into unified patient embeddings at hospital system scale. Moving from narrow, task-specific clinical models to holistic representations that predict disease, treatment response, and adverse events from one embedding is a fundamental shift for clinical AI.

📄 Read the paper.

🗓️ Events & Competitions

The best competitions, hackathons, and community challenges in AI x life sciences, curated weekly. Know something worth featuring? Reply and let us know.

📄Recap post: BIOMICS Hackathon | Feb 23-25

The BIOMICS hackathon at EMBL-EBI brought together computational biologists and software engineers from Portugal, Spain, Germany, and the UK for three days of building. Five challenge tracks covered everything from statistical tools to building software for a brand new microscopy technique from scratch.

Every team built something visual, reflecting a shift away from command-line-only workflows. This was also one of the first hackathons where AI coding agents like Claude Code were widely used across teams, and the difference in what could be achieved in three days was significant. One participant described these tools as an “exoskeleton” that amplifies existing ability.

BIOMICS is a EU Horizon-funded project twinning GIMM Lisbon with EMBL-EBI, CRG Barcelona, and ETH Zurich to strengthen biomedical data science training and collaboration. More events are planned throughout the year.

More upcoming events:

BioHackathon Europe 2026 | November 9-13, Barcelona

ELIXIR’s annual international bioinformatics hackathon, running since 2018. 160+ participants, five days of collaborative coding on open bioinformatics infrastructure and tools. The call for project proposals opens March 16 and closes April 15 - so if you want to lead a project, that’s your window.

Thanks for reading!

💬 Get involved

We’re always looking to grow our community. If you’d like to get involved, contribute ideas or share something you’re building, fill out this form or reach out to me directly.

Connect With Us

Have questions or suggestions? We'd love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

Keeping up with AI x life science news can get exhausting.

It’s scattered across LinkedIn, X, Substack, arXiv, Slack, newsletters... and you still somehow miss the things that actually matter. Too much noise, not enough signal.

We’re building something to fix that: a smarter, more powerful way to stay on top of what’s actually relevant to you.

But we want to build it with you, not just for you. Take 2 minutes to tell us what’s missing. What you share will directly shape what we build, and you’ll be the first to benefit from it.

Share your input

The failure rate in clinical development is well known but no less striking for it. Around 90% of clinical trials still fail. And buried inside that number is a problem the industry has not yet confronted: trials are still being designed on incomplete patient information, slowing recruitment, driving up attrition, and causing costly delays, ultimately putting medicines out of reach for the patients who need them most.

The industry spends an eyewatering $400 billion a year on that failure. And while the causes are multiple, a significant portion of that failure is traceable to decisions made at the protocol design stage, before a single patient is enrolled.

Trial teams are making high-stakes decisions about patient populations, eligibility criteria, endpoints, and recruitment strategies on a partial evidence base. Biology is well represented. But the patient’s life is not. The result is a systematic underestimation of the factors that actually determine recruitment and screening success, patient dropout, adherence, and the need for subsequent protocol amendments.

That’s the gap Luvida is building into.

We sat down with co-founders Hannah Amies and James Malone to understand what they’re working on. Hannah’s background spans biomedicine, consulting, head of product at BenevolentAI and epidemiology at Oxford. James brings computer science, bioinformatics, and a career spanning the European Bioinformatics Institute, his own acquired data curation company, and CTO roles at SciBite and Benevolent AI. Between them they cover both sides of the problem: the data science and the clinical domain knowledge.

🔴 The Problem

Clinical trial protocol design remains a surprisingly manual, consensus driven process for a field built on evidence. Getting a protocol ready involves assembling internal teams, bringing in external key opinion leaders, iterating over months, and more excel spreadsheets than are possible to manage. It can take up to 18 months, just for the design phase. And it’s not especially data-driven. “It’s very expert opinion driven,” James told us. “That can be very advantageous, you need that expertise. But it does mean biases creep in. Evidence is scattered across documents such as historical protocols, published literature, amendment documents, recruitment and on-trial data, and regulatory feedback. This is also a data problem.”

The result? Around 50% of trials end up requiring protocol amendments, averaging 3.3 per trial. Each one costs roughly $500,000 and burns at least three months waiting for regulatory sign-off. Do the maths across multiple trials per drug and you’re looking at hundreds of millions in lost on-patent revenue, and years of delay before a drug reaches a patient who needs it. “Most people working in the space are just doing it because they believe in getting good drugs into the right patients’ hands,” James said. That’s the real cost of a broken process.

📊 The Missing Data

The core issue is an incomplete patient picture. Clinical and biomedical data captures areas like biology, genotypic profiles, disease characteristics and prior treatment history. What it does not capture is the much broader set of variables that determine whether a given patient is recruited, adheres to treatment, or withdraws early.

The data exists, it is simply not being used. Hannah’s path to founding Luvida started not in a lab, but in Liverpool, implementing electronic patient records across three hospitals. That’s where she first noticed the gap: mountains of health data, almost entirely underleveraged. Epidemiology at Oxford sharpened the picture. “A lot of this stuff we have evidence for” she said, “but a lot of it is buried in research papers and not being leveraged at scale”

Luvida’s answer is what they call Electronic Life Records, a proprietary data layer that builds a more complete picture of the patient than clinical and biomedical data alone. It is that richer picture that changes what you can predict, and how accurately.

💡 The Idea: Expert in the Loop, Not AI in Charge

Luvida’s platform isn’t trying to replace clinical operations teams. James was clear: “We don’t want to come in and look like we’re replacing a clinical trial team of medical writers.” The goal is to speed up the parts of the job that involve synthesising signals from disparate, messy data sources, then hand that signal back to people who know what to do with it.

The platform works within a familiar authoring environment, think Google Docs-style interface, where AI-driven suggestions surface like comments: accept, reject, or edit. Every recommendation is tied directly to the evidence for review, which is used for decision-making and regulatory justification. There’s also a chat interface for question-answering. But the core of what Luvida does is pull together clinical and biomedical data, curated historical trial data, and patient lifestyle and behavioural data to flag where a trial is likely to fail before it starts.

Questions like: which populations are most likely to drop out? Where will recruitment stall? Are your eligibility criteria accidentally excluding a particular ethnic group? Does this background population have co-morbidities you haven’t accounted for? Is the evidence strong enough for regulatory review?

Crucially, it pairs AI pattern-finding with strict rule-based approaches. “AI is great at looking for big patterns,” James said, “but many modern models are trained to be helpful, which means they can sometimes infer things that aren’t really there.” So Luvida enforces rigour: everything is auditable, everything is traceable, and recommendations are evidence-based.

📢 Why It’s Different

For large pharma, Luvida gives teams the evidence base to navigate internal governance faster. Instead of one expert’s opinion against another’s, teams walk into meetings with data. For smaller biotechs who rely entirely on CROs and often feel, as Hannah put it, “pretty disempowered”, it’s even more significant. Luvida arms them with evidence to push back and ask better questions.

For rare diseases, where historical trial data is thin by definition, the platform can identify analogous disease areas and trials that work in adjacent spaces. “That’s something that AI is really good at identifying,” Hannah said. “The patterns. And that’s really where our models come into play.”

The early validation numbers are striking. When Luvida ran their models against past trials, feeding in initial protocols while blinding the system to the amendments that followed, it predicted 40- 50% of those subsequent amendments. In a prototype. The cost of each predicted amendment that doesn’t happen? Hundreds of thousands of dollars and three months of time.

🔮 The Future

Luvida is approximately a year old and is working with forward-thinking customers on live trial design programmes. Data is processed by indication to ensure data quality, a deliberate intention that keeps model outputs high quality and trustworthy. The platform already covers 500K+ trials, with 200K+ enriched with publications. From this, for example, we’ve already surfaced 87K+ known adherence risks.

The roadmap includes key opinion leaders contributing directly to the platform to streamline stakeholder management, and eventually patients too, to ensure the patient voice is brought in from the TPP, with protocols translated into plain language. Scenario modelling alternative trial designs enables pressure-testing of real-world on-trial risks before a trial even starts.

There’s also an API, so customers can plug Luvida into their own internal tooling and models, rather than adding another standalone system to an already crowded stack.

The platform is also designed with an eye on where regulation is heading, FDA diversity action plans and the NHS 10-year plan both point in the same direction: trials that are designed to reflect the real-world diversity of patient populations. Luvida’s richer patient picture makes that not just possible, but built in from the start.

The incomplete patient picture has been one of the industry’s most persistent and expensive blind spots for decades. Luvida is building the infrastructure to close it.

Get in touch

Luvida is headquartered in London and already partnering with CROs, pharma, and biotech across the EU and US. For a limited time, they’re offering exclusive value assessments. To see Luvida in action on one of your protocols, secure your spot at enquiries@luvida.co.uk.

You can also learn more at www.luvida.co.uk

Thanks for reading Kiin Bio Weekly!

💬 Get involved

We’re always looking to grow our community. If you’d like to get involved, contribute ideas or share something you’re building, fill out this form or reach out to me directly.

Share Kiin Bio Weekly

Subscribe now to stay at the forefront of AI in Life Science and keep up with this upcoming season of deep dives.

Connect With Us

Have questions on this or suggestions for our next deep dive? We’d love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

Keeping up with AI x life science news can get exhausting.

It’s scattered across LinkedIn, X, Substack, arXiv, Slack, newsletters... and you still somehow miss the things that actually matter. Too much noise, not enough signal.

We’re building something to fix that: a smarter, more powerful way to stay on top of what’s actually relevant to you.

But we want to build it with you, not just for you. Take 2 minutes to tell us what’s missing. What you share will directly shape what we build, and you’ll be the first to benefit from it.

Share your input

🇺🇸 Meet Us at Bio-IT, Boston

🤝 Our CEO Filippo and CTO Bogdan will be at Bio-IT World in Boston. If you are attending and would love to connect in person, reach out and let us know.

TmProt 1.0: Predicting Protein Melting Temperatures for Enzyme Discovery

🔬 Predicting protein melting temperatures (Tm) is critical for enzyme engineering: thermostable enzymes last longer and tolerate harsher conditions. But most AI predictors train on mass-spectrometry data from whole cells, which is fundamentally different from purified protein measurements. The Loschmidt Labs team found near-zero correlation (r = 0.05) between proteomics and biophysical Tm datasets for the same proteins.

TmProt 1.0 from Loschmidt Laboratories (Masaryk University) fixes this by training exclusively on biophysically measured melting temperatures. Built on ESM-2 fine-tuned with LoRA, it is now live on Hugging Face.

🧬 The team assembled ProMelt, a curated set of 45,441 proteins with experimental Tm values, validated across 5 independent biophysics test sets. TmProt outperforms TemBERTure, DeepSTABp, and SaProt across most benchmarks.

⚡ Absolute Tm regression remains hard, but TmProt excels where it matters most: ranking thermostable candidates. For classifying proteins with Tm at or above 60 degrees C, it achieves an AUC of 0.75, and is fully integrated into EnzymeMiner 2.0 for end-to-end enzyme mining.

🔬 Applications and Insights

1️⃣ Thermostable Enzyme Enrichment

Best used as a ranking tool to prioritise thermostable candidates from large sequence sets, making it practical for early-stage enzyme discovery campaigns.

2️⃣ Training Data Matters More Than Architecture

The near-zero correlation between proteomics and biophysical Tm explains why previous predictors underperformed. Curating ProMelt was the single biggest driver of improvement.

3️⃣ Lightweight and Accessible

LoRA fine-tuning keeps the model efficient enough to run as a free web server, lowering the barrier for labs without GPU infrastructure.

4️⃣ Integrated Discovery Pipeline

Full integration with EnzymeMiner 2.0 means researchers go from sequence database to stability-ranked shortlist in one workflow.

💡 Why This Is Cool

Data quality beating model complexity. The proteomics vs biophysical Tm disconnect explains years of underwhelming predictions. By fixing the data rather than stacking layers, TmProt delivers where enzyme engineers actually need it.

🌐 Try TmProt.

🌐 Try EnzymeMiner 2.0.

YuelDesign: Diffusion-Based Molecule Design in Flexible Protein Pockets

🔬 Proteins undergo conformational changes upon ligand binding, yet most deep learning generative models treat pockets as rigid, generating molecules for a single frozen conformation. The pocket shape in a crystal structure may not be the one your molecule actually binds.

YuelDesign from the Dokholyan lab at the University of Virginia jointly models both the pocket structure and ligand conformation, allowing protein and molecule to co-adapt during generation.

🧬 Two diffusion processes run simultaneously: an elucidated diffusion model (EDM) for 3D coordinates and a discrete denoising diffusion model (D3PM) for atom types. Both use E3former to maintain rotational and translational equivariance.

⚡ The result is molecules with favourable drug-likeness, low synthetic complexity, diverse functional groups, and docking energies comparable to native ligands. By letting the pocket breathe during generation, YuelDesign captures induced-fit effects that rigid methods miss.

🔬 Applications and Insights

1️⃣ Induced-Fit Drug Design

Jointly diffusing pocket and ligand explores conformational states that only emerge upon binding, capturing selectivity-driving dynamics rigid models cannot access.

2️⃣ Synthesisable and Drug-Like Output

Generated molecules score well on synthetic accessibility out of the box, reducing the gap between computational hits and what a medicinal chemist would actually make.

3️⃣ Diverse Chemical Exploration

D3PM atom-type diffusion produces varied functional groups rather than collapsing to a narrow series, giving broader chemical space coverage per run.

4️⃣ Beyond Rigid Docking

Treating flexibility explicitly moves generative design closer to real protein-ligand recognition, where both partners adjust shape upon binding.

💡 Why This Is Cool

The rigid-pocket assumption has limited structure-based design for decades. YuelDesign tackles it with a clean dual-diffusion architecture that still produces drug-like, synthetically accessible molecules. A meaningful step for generative molecular design.

📄 Read the paper.

💻 Try the code.

DISCO: AI-Driven Co-Design of Enzyme Sequence and Structure for Drug Discovery

🔬 Traditional enzyme design starts with a theozyme: a hand-crafted catalytic geometry based on detailed mechanistic knowledge. This works when you understand the mechanism, but limits you to chemistries where that knowledge exists.

DISCO (Diffusion for Sequence-Structure Co-design) from Caltech co-generates sequence and structure simultaneously using diffusion over discrete amino acids and continuous 3D coordinates, conditioned only on DFT-derived reactive intermediates. No predefined catalytic motifs needed.

🧬 Given a target chemistry, DISCO explores catalytic solutions freely, generating novel active sites and repurposing unrelated protein folds for catalysis. The model jointly optimises sequence and structure in a single pass, preserving their natural interdependence.

⚡ On the STUDIO-179 benchmark, DISCO generated the highest proportion of co-designable protein-ligand complexes for 178 of 179 cases. High-performing enzymes were found from just 90 designs, and for C(sp3)-H insertion, a poorly understood reaction, DISCO-designed enzymes matched variants from extensive directed evolution.

🔬 Applications and Insights

1️⃣ Bypassing Theozyme Design

Conditioning on reactive intermediates rather than predefined geometries lets DISCO tackle chemistries where mechanistic knowledge is incomplete or unavailable.

2️⃣ Evolvable by Design

Generated enzymes respond well to directed evolution, meaning computational design and experimental optimisation work as complementary stages.

3️⃣ Novel Folds for Novel Chemistry

DISCO repurposes unrelated protein scaffolds and generates new active sites, expanding the design space beyond known enzyme families.

4️⃣ Reduced Experimental Screening

Functional enzymes from just 90 designs dramatically reduces screening burden compared to traditional library approaches.

💡 Why This Is Cool

Enzyme design has always been bottlenecked by mechanistic knowledge. DISCO removes that constraint. Matching directed evolution on a poorly understood reaction from de novo designs suggests AI can now explore catalytic solutions humans would not think to try.

📄 Read the paper.

💻 Try the code.

🗓️ Events & Competitions

The best competitions, hackathons, and community challenges in AI x life sciences, curated weekly. Know something worth featuring? Reply and let us know.

More upcoming events:

Agentic Genomics: Hands-on with AI for Variant Interpretation and GWAS | April 22, Virtual

A free, two-session workshop co-hosted by Manuel Corpas and Segun Fatumo, running entirely in Google Colab. No setup, no cost, just a browser. Session one covers variant interpretation using Ensembl VEP, ClinVar, and ACMG criteria. Session two runs a full GWAS pipeline with ClawBio, including polygenic risk scores and locus fine-mapping. Built to remove barriers for researchers anywhere in the world.

Register (free): https://luma.com/vc98zeik

BioHackathon Europe 2026 | November 9-13, Barcelona

ELIXIR’s annual international bioinformatics hackathon, running since 2018. 160+ participants, five days of collaborative coding on open bioinformatics infrastructure and tools. The call for project proposals opens March 16 and closes April 15 - so if you want to lead a project, that’s your window.

Thanks for reading!

💬 Get involved

We’re always looking to grow our community. If you’d like to get involved, contribute ideas or share something you’re building, fill out this form or reach out to me directly.

Connect With Us

Have questions or suggestions? We'd love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

Keeping up with AI x life science news can get exhausting.

It’s scattered across LinkedIn, X, Substack, arXiv, Slack, newsletters... and you still somehow miss the things that actually matter. Too much noise, not enough signal.

We’re building something to fix that: a smarter, more powerful way to stay on top of what’s actually relevant to you.

But we want to build it with you, not just for you. Take 2 minutes to tell us what’s missing. What you share will directly shape what we build, and you’ll be the first to benefit from it.

Share your input

Today we’re having a look at Our Future Health, the UK’s largest health research programme, building a five‑million‑strong cohort to transform prevention, early detection and treatment. To understand how Our Future Health balances participant experience, scientific discovery, and data security, we sat down with James Siddle, a digital health consultant focusing on health insights within the programme.

🔴 The Problem

Large-scale health research programmes have shown how powerful population-level data can be for advancing prevention, diagnosis, and treatment. By linking biological samples, lifestyle information, and medical records, these initiatives have helped uncover genetic risk factors, guide new therapies, and inform public health policy.

Yet as the ambition and scale of such programmes grow, so do the challenges. Collecting and safeguarding highly sensitive data from millions of people demands exceptional security and governance. At the same time, public expectations are evolving; participants increasingly want to understand how their contribution supports discovery and how it might help their own health.

The central question for modern population health research is therefore how to build a system that serves both sides: one that enables cutting-edge science while ensuring every volunteer gains something valuable and personally meaningful in return.

💡 The Idea

Our Future Health brings together up to five million volunteers across the UK to help develop new ways to prevent, detect and treat diseases. Each participant contributes to creating one of the most detailed pictures of population health ever assembled, a resource designed to represent the full diversity of the UK and to power discoveries that improve everyone’s health.

Taking part involves completing health and lifestyle questionnaires, attending a short clinic appointment to have physical measurements taken, and providing a small blood sample. This information is securely linked with participants’ NHS records to build a rich, de-identified dataset that reflects real-world health across age, ethnicity, and region.

By combining this scale and depth of data with a focus on participant experience, Our Future Health aims to accelerate understanding of how different factors, such as genetics, environment, and behaviour interact to influence disease. Researchers from academia, the NHS, charities and industry can apply to access the data within a secure Trusted Research Environment, enabling them to explore new questions and develop better ways to predict, prevent and treat illness.

At its core, the programme is designed to create a virtuous cycle between public participation and research impact: people contribute information that helps science move forward, and the knowledge gained feeds back into better health outcomes for future generations.

🔬 Why It’s Different

Our Future Health is different to other health research programmes. Its design choices make it a discovery engine with unique advantages:

• Participant-first interface
  Our Future Health is developing an experience that prioritises participant clarity, ensuring that information is presented in a way that feels useful and easy to understand.

• Built-in recontact
  The programme plans to enable participants to be re-invited to future studies, creating opportunities for more dynamic research collaboration. As James noted, recontact requires “careful communication so people don’t feel alarmed,” a principle that guides every design decision.

• Scale & diversity
  With over two and a half million participants already enrolled, Our Future Health is the world’s largest health research programme of its kind and is on track to become the most diverse health cohort worldwide. Diversity across age, ethnicity, and geography ensures that the discoveries we make can benefit everyone.

• Security by design
  James emphasised that “information security is the top priority.” Our Future Health adheres to ISO 27001 standards and works closely with the UK’s National Cyber Security Centre. These measures are vital to maintaining participant trust and ensuring long-term sustainability.

• Active research underway
  Our Future Health is already enabling discovery. In June 2025, the first peer-reviewed study using its data was published in BMJ Mental Health by researchers at the University of Edinburgh. Analysing information from over 1.5 million participants, the team found that people with chronic inflammatory conditions may face nearly double the risk of mental-health issues compared with others.

This landmark paper marks the first of many studies set to use Our Future Health data to advance understanding of disease prevention, detection and treatment across a wide range of conditions.

🔮 The Future

As Our Future Health continues to grow toward its five-million-participant goal, its research value will expand exponentially. A recontactable, diverse population of this scale enables discoveries that could:

• Identify early markers of disease through long-term follow-up.

• Support targeted prevention strategies by highlighting who might benefit most from early interventions.

• Accelerate clinical trials by improving the speed and representativeness of recruitment.

For James and the Our Future Health team, success means giving participants something meaningful in return while advancing discoveries that improve health outcomes for everyone.

👨‍💻 Get in touch with James.

💻 Our Future Health Website.

🌐 Our Future Health on LinkedIn.

🤝 Get involved as a participant.

🔬 Access the data as a researcher.

Thanks for reading Kiin Bio Weekly!

💬 Get involved

We’re always looking to grow our community. If you’d like to get involved, contribute ideas or share something you’re building, fill out this form or reach out to me directly.

Share Kiin Bio Weekly

Subscribe now to stay at the forefront of AI in Life Science and keep up with this upcoming season of deep dives.

Connect With Us

Have questions on this or suggestions for our next deep dive? We’d love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

Keeping up with AI x life science news can get exhausting.

It’s scattered across LinkedIn, X, Substack, arXiv, Slack, newsletters... and you still somehow miss the things that actually matter. Too much noise, not enough signal.

We’re building something to fix that: a smarter, more powerful way to stay on top of what’s actually relevant to you.

But we want to build it with you, not just for you. Take 2 minutes to tell us what’s missing. What you share will directly shape what we build, and you’ll be the first to benefit from it.

Share your input

GATSBI: Improving Context-Aware Protein Embeddings Through Biologically Motivated Data Splits

🔬 Protein embeddings power everything from interaction prediction to functional annotation. But how we evaluate them matters just as much as how we build them. Random train/test splits let models cheat by memorizing well-connected proteins, inflating performance on benchmarks while hiding failures on the long tail of understudied proteins.

GATSBI (Graph Attention with Split-Boosted Inference) is a framework from Stanford that builds context-aware protein embeddings by integrating heterogeneous biological data, including PPIs from STRING, co-expression patterns, tissue-specific functional associations from HumanBase, and ESM-2 sequence representations, into a unified graph attention network.

🧬 The network covers 18,049 human proteins with 1.5M+ edges across three relationship types. Graph attention layers learn to weight different edge types and tissue contexts during message passing, producing embeddings that capture both local interactions and broader functional context.

⚡ The key insight is evaluation design. GATSBI introduces biologically motivated splits: edge splits (testing unseen relationships between known proteins) and node splits (testing entirely unseen proteins with less than 30% sequence identity). Across interaction prediction, function classification, and functional set prediction, GATSBI consistently outperforms PINNACLE, with the largest gains for understudied proteins (AUROC +0.259, AUPRC +0.290 on functional sets).

🔬 Applications and Insights

1️⃣ Plug-and-Play Protein Representations

Embeddings are pretrained and downloadable, making them drop-in features for downstream models predicting interactions, function, or pathway membership without retraining the graph.

2️⃣ Honest Benchmarking for Understudied Proteins

The node split reveals how models actually perform on proteins with limited prior evidence, exposing a critical gap that random splits conceal in current benchmarking.

3️⃣ Biologically Plausible “False Positives”

High-confidence predictions flagged as errors turned out to reflect real but unannotated relationships, suggesting the model captures genuine biology beyond current annotations.

4️⃣ Boosting the Long Tail of the Proteome

The heterogeneous graph adds the most information for low-degree proteins, precisely the understudied targets where computational predictions matter most for discovery.

💡 Why This Is Cool

The lesson here goes beyond protein embeddings. How you split your data determines what your benchmark actually measures. GATSBI shows that when you evaluate properly, the gap between methods changes dramatically, and models that look equivalent on well-studied proteins diverge sharply on the understudied ones that matter for real discovery.

📄 Read the paper.

💻 Try the code.

ProtiCelli: Generative Machine Learning Unlocks the First Proteome-Wide Image of Human Cells

🔬 Current imaging technologies can visualize tens of proteins simultaneously. A single human cell contains thousands. This gap means we still lack a complete picture of how the proteome is spatially organized, and brute-force experimental coverage is not practical at this scale.

ProtiCelli is a deep generative model from Stanford and KTH, trained on 1.23 million immunofluorescence images from the Human Protein Atlas, that simulates microscopy images for 12,800 human proteins using just three cellular landmark stains as input (nucleus, microtubules, ER).

🧬 Given only the landmark channels, the model predicts what the protein channel would look like, effectively hallucinating biologically accurate fluorescence patterns. It generalizes to unseen cell types and drug perturbations, preserves hierarchical subcellular organization, and even recapitulates known protein-protein interactions from the spatial patterns alone.

⚡ The team generated Proteome2Cell: 30.7 million simulated images representing 2,400 virtual cells across 12 human cell lines, now integrated into the Human Protein Atlas. The model can also infer drug-induced changes in protein localization from cell morphology alone, without ever seeing the drug-treated protein images during training.

🔬 Applications and Insights

1️⃣ Virtual Proteome-Scale Experiments

Enables imaging experiments that would take years and millions of dollars to run physically, compressing exploration of protein spatial organization into a single model inference.

2️⃣ Drug Mechanism-of-Action Discovery

Predicts how compounds alter protein localization from cell morphology alone, without staining for every target, potentially accelerating early-stage perturbation screening.

3️⃣ Orthogonal Interaction Signal

Predicted co-localization patterns correlate with known protein-protein interactions, providing a spatial signal that complements sequence-based interaction predictions.

4️⃣ A Public Resource for the Community

Proteome2Cell (30.7M images across 12 cell lines) is freely available through the Human Protein Atlas, giving researchers a shared baseline for cell biology and drug discovery.

💡 Why This Is Cool

This is the microscopy equivalent of protein structure prediction. Just as AlphaFold gave us predicted structures for proteins we had not crystallized, ProtiCelli gives us predicted images for proteins we have not stained. The fact that drug-induced changes emerge from morphology alone suggests the model has learned something genuinely deep about the relationship between cell shape and protein organisation.

📄 Read the paper.

💻 Try the code.

StriMap: A Structure-Informed Deep Learning Framework for TCR-Peptide-HLA Interactions

🔬 T cell receptor (TCR) recognition of peptide-HLA complexes drives adaptive immunity, from fighting infections to rejecting tumors to triggering autoimmunity. But predicting which TCR will bind which peptide-HLA is notoriously difficult: the interaction interface is flexible, the sequence space is vast, and training data is sparse.

StriMap is a unified deep learning framework from the Xavier lab at the Broad Institute and MIT that integrates physicochemical features, sequence context, and structural information at the recognition interface to predict TCR-peptide-HLA binding.

🧬 Rather than relying on sequence alone, StriMap models the 3D contact geometry between TCR CDR loops and the peptide-HLA surface. By combining structural priors with learned sequence representations, it captures binding patterns that pure sequence models miss, achieving state-of-the-art performance with improved generalizability across diverse datasets.

⚡ The real validation came from application: the team screened 13 million peptides from 43,241 bacterial proteins to find molecular mimics relevant to ankylosing spondylitis (AS). Top candidates were experimentally validated to activate T cells expressing an AS-associated TCR, and one peptide showed enrichment in inflammatory bowel disease patients, suggesting shared microbial triggers between AS and IBD.

🔬 Applications and Insights

1️⃣ Generalizable TCR Binding Prediction

Structure-informed features at the binding interface improve generalization beyond the peptides and HLAs seen during training, breaking a critical bottleneck for TCR prediction models.

2️⃣ Computational-to-Experimental Discovery Pipeline

The AS screening demonstrates a complete workflow: computational prediction of 13M peptides narrowed to candidates that were experimentally validated to activate disease-relevant T cells.

3️⃣ Cross-Disease Molecular Mimicry

A top validated microbial peptide showed enrichment in IBD patients, supporting the hypothesis that shared bacterial triggers drive immune cross-reactivity between AS and IBD.

4️⃣ Dual Application in Cancer and Autoimmunity

The framework applies to both cancer neoantigen prediction and autoimmune antigen discovery, bridging two major areas of immunotherapy.

💡 Why This Is Cool

TCR-antigen prediction has been one of the stubbornest problems in computational immunology. Most models fail to generalize beyond their training peptides. By bringing structural information into the prediction, StriMap breaks out of that limitation, and the AS/IBD discovery shows this is not just a benchmark improvement but a tool that finds real biology.

📄 Read the paper

💻 Try the code.

🗓️ Events & Competitions

The best competitions, hackathons, and community challenges in AI x life sciences, curated weekly. Know something worth featuring? Reply and let us know.

More upcoming events:

Agentic Genomics: Hands-on with AI for Variant Interpretation and GWAS | April 22, Virtual

A free, two-session workshop co-hosted by Manuel Corpas and Segun Fatumo, running entirely in Google Colab. No setup, no cost, just a browser. Session one covers variant interpretation using Ensembl VEP, ClinVar, and ACMG criteria. Session two runs a full GWAS pipeline with ClawBio, including polygenic risk scores and locus fine-mapping. Built to remove barriers for researchers anywhere in the world.

Register (free): https://luma.com/vc98zeik

BioHackathon Europe 2026 | November 9-13, Barcelona

ELIXIR’s annual international bioinformatics hackathon, running since 2018. 160+ participants, five days of collaborative coding on open bioinformatics infrastructure and tools. The call for project proposals opens March 16 and closes April 15 - so if you want to lead a project, that’s your window.

Thanks for reading!

💬 Get involved

We’re always looking to grow our community. If you’d like to get involved, contribute ideas or share something you’re building, fill out this form or reach out to me directly.

Connect With Us

Have questions or suggestions? We'd love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website

For decades, designing a new protein meant years of directed evolution, rational engineering, and a lot of luck. You started from something nature already made and slowly nudged it toward what you wanted. The success rate was low. The timelines were long.

That’s changing fast. A wave of AI models, led by tools like RFDiffusion, ProteinMPNN, and AlphaFold, has opened up a fundamentally different approach: designing proteins from scratch, computationally, and getting functional molecules on the first try. In the last two years, AI-designed proteins have matched or outperformed naturally evolved ones in binding affinity, stability, and specificity, sometimes by significant margins.

This primer covers what’s actually happening, how the key methods work, and where the field is headed.

Keeping up with AI x life science news can get exhausting.

It’s scattered across LinkedIn, X, Substack, arXiv, Slack, newsletters... and you still somehow miss the things that actually matter. Too much noise, not enough signal.

We’re building something to fix that: a smarter, more powerful way to stay on top of what’s actually relevant to you.

But we want to build it with you, not just for you. Take 2 minutes to tell us what’s missing. What you share will directly shape what we build, and you’ll be the first to benefit from it.

Share your input

🔬 From Prediction to Design

The story starts with structure prediction. AlphaFold, released by DeepMind in 2020, solved a 50-year-old problem: predicting a protein’s 3D structure from its amino acid sequence. That was transformative for understanding biology, but it didn’t directly design new proteins. It told you what a sequence would fold into, not what sequence would give you the fold you wanted.

The leap to design required inverting the problem. Instead of sequence → structure, the question became: what sequence would fold into a structure that does what I need?

That inversion is what the current generation of tools enables.

🧩 The Key Methods

There are three categories of AI tools driving protein design today, and they work best in combination.

Structure generation creates new protein backbone shapes, the 3D scaffold. The breakthrough here is RFDiffusion, developed by David Baker’s lab at the University of Washington. It uses diffusion models (the same class of generative AI behind image tools like DALL-E) applied to 3D coordinates. You specify what you want, a protein that binds a particular target, wraps around a small molecule, or presents a specific functional site, and the model generates backbone structures that satisfy those constraints. It’s designing architectures that evolution never explored.

Sequence design fills in the amino acid sequence for a given backbone. ProteinMPNN, also from Baker’s lab, takes a 3D structure and predicts which amino acid sequences will fold into it stably. This is the bridge between a computational shape and something you can actually synthesise and test. It recovers native-like sequences with high accuracy and, critically, produces sequences that fold and function when tested experimentally.

Structure prediction closes the loop. AlphaFold (and its open-source successor ESMFold from Meta) validates the designs by predicting whether the designed sequence will actually fold into the intended structure. If the predicted fold matches the designed backbone, confidence is high. If it doesn’t, you iterate.

The typical workflow today: RFDiffusion generates a backbone → ProteinMPNN designs sequences for it → AlphaFold confirms the fold → the best candidates go to the lab.

The AI protein design workflow: a diffusion model generates novel backbone structures, a sequence design model fills in amino acids, and a structure prediction model validates the fold before experimental testing.

⚙️ What’s Actually Working

The results from the last 18 months have been striking.

De novo binders, proteins designed from scratch to bind a specific target, are now routinely achieving nanomolar affinity on the first experimental round, without any optimisation. A 2024 study from Baker’s lab designed binders against a panel of therapeutic targets, including influenza and SARS-CoV-2, with success rates that would have been unthinkable five years ago.

Protein design competitions are providing independent validation. Adaptyv Bio, a cloud lab for protein designers based in Lausanne, ran an open EGFR binder competition in 2024 that benchmarked AI design methods head-to-head with standardised experimental testing. The results showed a 5x improvement in design success rates compared to earlier approaches, with some AI-designed binders outperforming clinical antibodies.

Stability is also improving. AI-designed proteins are increasingly more thermostable than their natural counterparts. They can be engineered to withstand higher temperatures and harsher conditions, which matters enormously for manufacturing and storage.

🧪 The Validation Bottleneck

Here’s the catch: designing a protein computationally takes hours. Testing it experimentally still takes weeks to months.

The field can now generate thousands of candidate designs per day. But each one needs to be synthesised, expressed, purified, and assayed to know if it actually works. That wet-lab step is the bottleneck, and it’s where a lot of promising computational designs die, not because the design was wrong, but because the testing pipeline can’t keep up.

This is driving a new category of infrastructure: automated, high-throughput protein testing platforms that can validate designs at the speed AI generates them. The goal is a closed loop, design, test, learn, redesign, running continuously with minimal manual intervention. We’ll be exploring this challenge in an upcoming deep dive with Adaptyv Bio, a cloud lab purpose-built for AI protein design validation. Stay tuned.

Until that loop is fully closed, the practical throughput of AI protein design is limited not by the models but by the experiments.

📊 Beyond Binders

Binding is the easiest thing to design for, because the objective is clear: does this protein stick to that target? But the field is pushing into harder problems.

Enzyme design, creating proteins that catalyse specific chemical reactions, is significantly more challenging because function depends on precise atomic arrangements in the active site, not just overall shape. Early results are promising but success rates are lower than for binders.

Multi-state design aims to create proteins that switch between conformations, molecular machines that respond to signals. This requires the model to optimise for multiple structures simultaneously, a much harder optimisation problem.

Symmetric assemblies, protein cages, rings, and lattices, are being designed for drug delivery, vaccine design, and materials science. RFDiffusion has demonstrated the ability to generate novel symmetric architectures that self-assemble when tested experimentally.

🔮 Where This Is Going

Three trends to watch.

Generative models are getting multimodal. The next generation of design tools will jointly generate structure and sequence, rather than treating them as separate steps. Models that can reason about structure, sequence, dynamics, and function simultaneously will produce better designs faster.

The data flywheel is spinning up. Every experimentally tested design, whether it works or not, generates training data that makes the next round of models better. Open repositories for protein design data are accelerating this. The more designs get tested, the faster the models improve.

The design-test loop is tightening. As automated testing platforms scale, the gap between computational design and experimental validation will shrink. The long-term vision is protein design on demand: specify the function you want, get a validated molecule back in days rather than months.

We’re still early. Most AI-designed proteins are relatively simple, single-domain binders tested in controlled settings. The gap between designing a protein that binds a target in a tube and one that works as a drug in a patient remains enormous. But the trajectory is clear: the tools are getting better, faster, and more accessible. And the proteins they’re producing are starting to work.

💬 Want to be featured in Kiin Bio Weekly?

Each issue we speak directly with researchers, scientists, and builders working at the frontier of AI in life sciences. If you're working on something in this space and think it would resonate with our community, I'd love to hear from you - fill out this form or reach out to me directly.

Found this useful? Forward it to a colleague in drug discovery or protein design - it's the best way to help the newsletter grow.

Share Kiin Bio Weekly

Subscribe now to stay at the forefront of AI in Life Science. Every week: primers, deep dives, and direct conversations with the people building the field.

Subscribe now

Connect With Us

Have questions on this or suggestions for our next deep dive? We’d love to hear from you!

📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website