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

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

G4REP: RNA G-quadruplex-binding protein prediction across the human proteome

🧬 RNA G-quadruplex-binding proteins regulate mRNA processing, localisation, and stress responses - but experimental detection alone can’t scale. G4REP maps them across the entire human proteome.

🔬 RNA G-quadruplexes act as post-transcriptional regulatory hubs, but identifying which proteins bind them is experimentally challenging. Classical approaches rely on known binding domains like zinc fingers, missing a wide range of RG4-interacting proteins.

Researchers at the Università degli Studi di Modena e Reggio Emilia and Sapienza Università di Roma introduce G4REP, combining ESM-2 embeddings with LSTM neural networks to predict RG4-binding proteins at proteome scale.

🧬 G4REP analyses protein sequences for RG4 binding features: arginine-glycine-rich motifs, intrinsically disordered regions, and aromatic residues, identifying the short flexible motifs through which interactions typically occur.

⚡ ~85% accuracy and AUROC of 0.91. Over 2000 candidate RG4-binding proteins identified across the human proteome, including 552 high-confidence hits.

🔬Applications and Insights

1️⃣ Identifying Therapeutic Targets

RG4-binding proteins are enriched in stress granules linked to cancer and neurodegeneration. G4REP identifies candidates with ~85% accuracy, enabling reliable prioritisation of targets influencing RNA stability and translation.

2️⃣ Expanding RNA-Binding Protein Networks

G4REP identified 2000+ candidate proteins including poorly characterised FAM families, expanding RNA regulation well beyond classical binding domains.

3️⃣ Predicting Functional Binding Regions

G4REP pinpoints binding sites within proteins using a disorder-weighted residue score, highlighting short RGG-rich motifs as primary interaction sites.

4️⃣ Understanding Cellular Stress Responses

552 high-confidence RG4-binding proteins localise to stress granules, supporting a coordinating role for RG4 interactions in RNA metabolism during cellular stress.

💡 Why This Is Cool

G4REP opens up RG4 biology at a scale experimental methods alone cannot reach. It expands our map of RNA-protein interactions and surfaces previously hidden regulators and therapeutic candidates across the human proteome.

📖 Read the paper

💻 Code: https://lnkd.in/gYTkHmRd

evedesign: accessible biosequence design with a unified framework

🧬 Protein design has dozens of ML models. None of them talk to each other. Every real-world project still requires custom glue code, reformatting, and one-off pipelines. evedesign changes that.

🔬 The design problems that matter most in protein engineering - conditional design under real-world constraints, multi-objective optimisation, and iterative lab-in-the-loop workflows - demand flexible, composable infrastructure that no single tool provides. Current ML methods are rarely interoperable and remain inaccessible to non-experts.

Researchers at Harvard Medical School, Wellcome Sanger Institute, University of Cambridge, Seoul National University, Broad Institute, and collaborators built evedesign, a unified open-source framework that formalises conditional biosequence design in a method-agnostic way.

🧬 evedesign defines three composable operations - Generate, Score, and Transform - that work across any model type (MSA-based, LLM, inverse folding, de novo 3D). A standardised multi-level instance representation (sequence, embedding, and 3D structure simultaneously) lets outputs from one model feed directly into another without reformatting.

⚡ Supports multi-objective optimisation, supervised and unsupervised model integration, and lab-in-the-loop iteration from the ground up. Interactive web interface at evedesign.bio takes users from target protein to orderable DNA. Demonstrated in antibody engineering, enzyme design, and natural enzyme discovery. MIT-licensed, FAIR-compliant, and self-hostable.

🔬Applications & Insights

1️⃣ Antibody Engineering Conditional

CDR design subject to multiple constraints (thermostability, deimmunisation, binding) using composable multi-objective pipelines - no custom code required.

2️⃣ Enzyme Design

Combine generative models with scoring functions from different method families in a single workflow, iterating with experimental data as it comes in.

3️⃣ Lab-in-the-Loop Workflows

Declarative, serialisable pipelines that can halt, incorporate new experimental results, and resume - built for iterative design rounds rather than one-shot prediction.

4️⃣ Accessibility for Non-Computational

Researchers End-to-end web interface makes ML-driven protein design accessible to experimentalists without requiring programming or model-specific expertise.

💡 Why This Is Cool

The bottleneck in protein engineering isn’t any individual model - it’s connecting them. evedesign is the first open-source framework that treats this as a first-class problem: standardised interfaces, composable workflows, and lab-in-the-loop iteration built in from the start. That’s infrastructure the field has needed for years.

📖 Read the paper

💻 Website

Peptide-protein docking: from physics-based models to generative intelligence

🔬 Peptide therapeutics are one of the fastest-growing drug modalities - they target flat, extended protein surfaces that small molecules can’t reach. But computationally predicting how a peptide binds its target remains hard: peptides are flexible, often disordered, and fold upon binding. Classical docking methods struggle with all three.

Researchers at Purdue University and Korea University College of Medicine review the full landscape of peptide-protein docking, from traditional physics-based approaches to the latest deep learning methods, covering 25+ tools across three generations.

🧬 The review organises methods into three categories: binding site predictors that guide or filter docking models, AlphaFold-based protocols for peptide-protein co-folding and refinement, and deep generative models (diffusion-based) that sample peptide conformations conditioned on a target structure.

⚡ Diffusion models like RAPiDock and DiffPepDock are emerging as the most promising direction, handling peptide flexibility natively. AlphaFold-based methods work well for structured peptides but struggle with disordered ones. Major remaining gaps: limited training data, weak performance on long peptides, and almost no methods handle chemically modified peptides.

🔬Applications & Insights

1️⃣ Peptide Drug Design

Maps the full computational toolkit for predicting peptide-protein binding modes - essential for designing peptide therapeutics against previously undruggable protein surfaces.

2️⃣ Masking Peptide Engineering

Covers tools applicable to masking peptides that block drug binding sites until protease activation at the tumour - a growing immunotherapy strategy.

3️⃣ Practical Method Selection

Table 1 catalogues 25+ methods with years, descriptions, and code availability links - a ready-made decision guide for choosing the right docking approach.

4️⃣ Open Challenges

Identifies the three biggest gaps: training data scarcity, long/disordered peptide performance, and chemical modification handling - a roadmap for future method development.

💡 Why This Is Cool

This is the review the peptide docking field needed. It doesn’t just catalogue methods - it explains when each approach works, when it breaks, and what’s missing. If you’re working on peptide therapeutics, this is your decision guide for computational docking in 2026.

📖 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.

🏗️ Hackathon Highlight

The Elnora x Monomer Bio AI scientist hackathon brought together 43 builders for 24 hours of agent-driven lab automation. The winning team connected Elnora’s protocol generation agent to a robotic arm that autonomously imaged cells, assessed confluency at 70%, generated a splitting protocol, executed it, and re-imaged to confirm results - a fully closed AI-to-robot lab loop built from scratch overnight. 6 out of 7 teams incorporated Elnora into their workflows.

Full recap from CEO Carmen Kivisild.

Recap on Edinburgh BioHackathon

Last weekend, Edinburgh hosted its first ever biohackathon, and the numbers speak for themselves.

🎯 350 applications. 110 selected participants. Over 95% attendance rate, which for a free hackathon is exceptional (typical rates sit around 40-50%).

Organised entirely from scratch by a volunteer team led by Xi Yang AMRSC and Yen Peng Chew, PhD, AFHEA at the The University of Edinburgh, BioHackathon Edinburgh (PRIMED) brought together students, postdocs and clinical researchers from across Scotland: 52% from The University of Edinburgh, 48% from University of Dundee, University of St Andrews, University of Stirling, University of Glasgow and beyond.

Seven challenges spanned three tracks: academic, industry and non-coder. The two standouts were the bio-business and Pacifico Biolabs (Genome-Scale Metabolic Modelling Tool) track, which saw the highest participation with 8 and 7 teams respectively.

🔬 One project that caught everyone's attention: FilamentTracker. A team built a fully functional web tool with its own custom domain that detects and tracks protein filaments in microscopy images, showing filament length and area over time. In 48 hours. Judges flagged it as potentially publishable if stress-tested across other organisms.

What made this event different was the deliberate push for interdisciplinary collaboration. Teams were encouraged to mix engineers, biologists, data scientists and non-coders. Not everyone landed in a mixed team, but the feedback was clear: those who did got the most out of it.

All submissions are open source on DevPost (https://lnkd.in/gjpQwXXR). Next up from the organisers: peer-led company creation workshops to help winning teams take their projects further.

🚀 Scotland's first biohackathon, built from a blank Google Drive folder by an unpaid volunteer team. If this is what version one looks like, version two is going to be something special!

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

CellVoyager: An AI agent for autonomous biological data analysis

🔬 Most AI tools for single-cell analysis execute the analyses a user asks for. CellVoyager does the opposite: given a processed scRNA-seq dataset and prior analyses from a published paper, it autonomously proposes and executes novel analytical directions that build on existing work.

Stanford University’s Zou Lab built CellVoyager on LLMs running within a fixed Jupyter environment. It generates iterative “exploration blueprints” - self-critiqued analytical plans executed step-by-step with automatic code fixing and VLM-based result interpretation.

🧬 CellVoyager conditions each new hypothesis on what has already been attempted, preventing redundancy. It generates code, runs it, interprets outputs via a vision-language model, and updates the exploration plan accordingly.

⚡ Outperforms GPT-4o baseline by 16% (micro-averaged, p<0.01) and 19.33% (macro-averaged, p<0.001) on CellBench - 50 published scRNA-seq studies, 483 analyses. Discovered CD8+ T cells in COVID-19 are more primed for pyroptosis and a link between transcriptional noise and aging in the brain’s subventricular zone - both absent from the original papers.

🔬 Applications & Insights

1️⃣ Hypothesis Generation from Published Data

Proposes novel analytical directions beyond those in the original paper - without new experiments or new data.

2️⃣ COVID-19 Immunology

Found CD8+ T cells in COVID-19 patients are more primed for pyroptosis - a mechanistic insight absent from the original study.

3️⃣ Brain Aging

Discovered a link between transcriptional noise and aging in the subventricular zone, validated as novel by the original authors.

4️⃣ Collaborative Research

Acceleration Ingests what has been done, explores what hasn’t, and produces interpretable reports for expert review.

💡 Why This Is Cool

Most scRNA-seq datasets have more biology in them than the original analysis captured. CellVoyager treats that gap as an opportunity. The benchmark measures whether the agent predicts what a paper actually analysed - a much harder standard than held-out prediction tasks.

📖 Read the paper

💻 GitHub

MacroGuide: Topological guidance for macrocycle generation

🔬 Macrocycles offer superior selectivity against difficult drug targets, but generative models almost never produce them. The core problem is topological: ring closure is a global structural constraint that local generative approaches can’t enforce.

University of Oxford, AITHYRA, ENS Ulm, and TU Wien introduce MacroGuide, a training-free diffusion guidance mechanism that steers any pretrained molecular generative model toward macrocycles using Persistent Homology.

🧬 At each denoising step, MacroGuide constructs a Vietoris-Rips complex from atomic positions and computes gradients from three topological objectives: cycle size (H1 death), cycle connectivity (H1 birth), and molecule connectivity (H0 death). These steer the score function toward ring-forming structures without modifying the base model.

⚡ Macrocycle generation rate from 1% to 99% on pretrained diffusion models. Matches or exceeds SOTA on chemical validity, diversity, PoseBusters checks, and pharmacophore satisfaction. Demonstrated in unconditional and protein pocket-conditioned settings including bicyclic structures.

🔬 Applications & Insights

1️⃣ De Novo Macrocycle Design

Not limited to cyclic peptides or linear scaffolds - enables arbitrary ring architectures without requiring known linear equivalents.

2️⃣ Plug-and-Play Compatibility

Training-free. Plugs into any pretrained diffusion model without retraining or fine-tuning.

3️⃣ Structure-Based Drug Design

Generates macrocycles conditioned on protein binding pockets, including bicyclic structures, for structure-based drug design campaigns.

4️⃣ Difficult Target Access

Macrocycles bind protein surfaces and allosteric sites unreachable by small molecules. MacroGuide makes these architectures designable at scale.

💡 Why This Is Cool

Macrocycles hit targets that small molecules can’t - but designing them required peptide chemistry or handcrafted scaffolds. MacroGuide removes both constraints as a plug-in for any diffusion-based molecular generation workflow. No retraining, no approximation.

📖 Read the paper

BioReason-Pro: Multimodal biological reasoning for protein function prediction

🔬 Standard protein function prediction treats GO annotation as classification - it gives a label but not a reason. Expert biologists reason across sequence, structure, domains, evolution, and interaction networks. BioReason-Pro was built to do the same.

The Arc Institute, University of Toronto, Vector Institute, and Stanford built BioReason-Pro, the first multimodal reasoning LLM for protein function prediction, generating structured step-by-step biological reasoning traces.

🧬 BioReason-Pro integrates ESM3 residue embeddings, GO graph structure, STRING protein interactions, and InterPro domain annotations. GO-GPT, an autoregressive transformer, provides sequential GO term predictions as context. Fine-tuned on 133K+ proteins via SFT on GPT-5-generated reasoning traces, then optimised with RL using GO prediction accuracy as reward.

⚡ GO-GPT achieves F_max = 0.70 (best-of-10), outperforming InterLabelGO+ and ProtBoost. Human protein experts preferred BioReason-Pro over ground truth UniProt annotations in 79% of cases. Attention for DNA-binding proteins aligns to exact binding residues (AUROC 0.81, 2.8x fold-enrichment) - without structural input.

🔬 Applications & Insights

1️⃣ Protein Annotation at Scale

Applied to 240,000+ proteins including the Human Protein Atlas, covering the vast majority with no experimental annotations.

2️⃣ Binding Partner Prediction

Zero-shot identification with attention aligning to cryo-EM-resolved contact residues - structural validation without structural input.

3️⃣ Beyond Homology

Integrates multiple evidence types to override misleading superfamily annotations, beyond what homology transfer alone achieves.

4️⃣ Understudied Proteins

Robust on proteins with low training similarity - relevant for viral proteins, rare organisms, and novel therapeutic targets.

💡 Why This Is Cool

Most protein function tools give you a label. BioReason-Pro gives you a reason. Human experts preferred its annotations over UniProt ground truth in 79% of cases - the model’s reasoning is considered more credible than the existing gold standard. That’s a meaningful bar.

📖 Read the paper

💻 GitHub

🗓️ 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.

Check this section next week for a recap with the organisers of BioHackathon Edinburgh 2026!

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

Eubiota: Modular agentic AI for autonomous discovery in the gut microbiome

🔬 Microbiome research generates vast genomic, metabolic, and clinical data, but connecting it to mechanistic discoveries requires navigating thousands of tools, databases, and experimental protocols. Most AI systems handle isolated steps - none handle the full scientific workflow.

Stanford University and Chan Zuckerberg Biohub built Eubiota, an open-source agentic AI framework with specialised agents for planning, execution, verification, and synthesis, connected through shared memory and reinforcement-learned reasoning.

🧬 Eubiota coordinates domain-specific tools across the microbiome research workflow: gene screening, metabolite analysis, experimental design, and literature synthesis. Reinforcement learning optimises multi-turn reasoning, allowing the system to refine hypotheses across iterative experimental cycles.

⚡ 87.7% accuracy on microbiome benchmarks, exceeding GPT-5.1 by 10.4%. Screened nearly 2,000 bacterial genes to identify the uvr-ruv DNA repair axis as a fitness determinant under inflammatory stress. Designed a four-strain consortium that reduced colitis severity in mice. Generated a commensal-sparing antibiotic cocktail and discovered diet-associated metabolites suppressing NF-kB signalling.

🔬 Applications & Insights

1️⃣ Autonomous Gene Screening

Screened ~2,000 bacterial genes to identify mechanisms driving fitness under gut inflammation - a scale and speed not achievable with manual experimental design.

2️⃣ Therapeutic Consortium

Design Designed a four-strain bacterial consortium that reduced colitis severity in a mouse model, demonstrating the framework’s ability to move from discovery to experimental validation.

3️⃣ Antibiotic Development

Generated a commensal-sparing antibiotic cocktail - targeting pathogens while preserving the healthy microbiome, a longstanding challenge in antimicrobial therapy.

4️⃣ Metabolite-Immune Crosstalk

Identified diet-associated metabolites that suppress NF-kB signalling, opening new angles for dietary and metabolic interventions in inflammatory disease.

💡 Why This Is Cool

Eubiota changes what’s possible in microbiome research by removing the coordination bottleneck. The most valuable finding here isn’t any single discovery - it’s that an AI system can now run the iterative hypothesis-experiment-synthesis loop that normally takes a large research team years. That’s a different category of tool.

📖 Read the paper.

💻 Try the code.

Fold-CP: A context parallelism framework for biomolecular modelling

🔬 AlphaFold-style models use a pairwise representation that scales quadratically with sequence length, capping useful inference at a few thousand residues on a single GPU. Over 70% of characterised mammalian protein complexes in the CORUM database exceed this limit, leaving most disease-relevant assemblies structurally inaccessible.

NVIDIA BioNeMo Fold-CP distributes the inference and training of co-folding models across multiple GPUs, using the open-source Boltz architecture as a reference.

🧬 A 2D device mesh tiles the N×N pair representation across GPUs so per-device memory scales as O(N²/P). Custom distributed algorithms handle triangle attention, triangle multiplication, and window-batched local attention without cropping, chunking, or approximation.

⚡ Structure prediction of assemblies exceeding 30,000 residues on 64 NVIDIA B300 GPUs. Accuracy parity with single-device baselines (R=0.97, median lDDT difference 0.0007). Scored 93% of the CORUM mammalian complex database, up from <30% accessible on standard models.

🔬 Applications & Insights

1️⃣ Mega-Complex Structure Prediction

Assemblies exceeding 30,000 residues predicted on 64 NVIDIA B300 GPUs. Over 70% of CORUM mammalian complexes now structurally accessible, up from <30% on standard single-GPU models.

2️⃣ Disease-Relevant Targets

PI4KA, a 3,605-residue complex linked to inflammatory bowel disease, predicted with TM-score 0.83. Previously inaccessible to co-folding models on standard hardware.

3️⃣ Accuracy at Scale

Prediction quality matches single-device baselines (R=0.97, median lDDT difference 0.0007), confirming parallelisation does not degrade structure quality.

4️⃣ Memory Scaling without Approximation O(N²/P) memory scaling across P GPUs. No cropping, chunking, or approximation. Custom distributed triangle attention and multiplication preserve full-model accuracy.

💡 Why This Is Cool

The bottleneck in structural biology has shifted from sequence to scale. Fold-CP doesn’t approximate around the quadratic memory wall, it removes it. The 93% CORUM coverage figure is striking: most disease-relevant complexes were structurally inaccessible before this. That changes what’s tractable for drug discovery.

📖 Read the paper

💻Try the code.

X-Cell: Scaling causal perturbation prediction across diverse cellular contexts

🔬 Most transcriptomic perturbation models train on a single cell type and fail elsewhere. Xaira Therapeutics built both the dataset and the model to change that: X-Atlas/Pisces, 25.6 million perturbed single-cell transcriptomes across 16 contexts, and X-Cell, a diffusion language model that predicts genome-wide perturbation responses.

🧬 X-Cell refines predictions by iteratively denoising control-to-perturbed transitions, conditioning on five biological priors via cross-attention: STRING, ESM-2, DepMap, GenePT, and CellPairing.

⚡ Up to 5-fold improvement over SOTA (Pearson Delta 0.51 vs STATE’s 0.10). Zero-shot T cell inactivation prediction in Jurkat cells. X-Cell-Ultra at 4.9 billion parameters is the first perturbation model to follow LLM power-law scaling. Zero-shot generalisation to iPSC melanocytes and primary CD4+ T cells.

🔬 Applications & Insights

1️⃣ Drug Target Prioritisation

Predict transcriptomic effects of gene knockouts across diverse cell types, reducing reliance on costly per-indication Perturb-seq screens.

2️⃣ Immunotherapy

Zero-shot prediction of T cell inactivating perturbations - identifying knockouts that shift activated T cells toward a resting state for autoimmune and immunotherapy applications.

3️⃣ Drug Response Prediction

X-Cell transfers from genetic to chemical perturbations (Pearson Delta 0.31 vs STATE’s 0.22 on Tahoe-100M drug data), opening paths to in silico drug screening.

4️⃣ Perturbation Biology Scales Like LLMs

X-Cell-Ultra follows LLM-style power-law scaling - the first evidence that larger models and more data consistently improve perturbation prediction.

💡 Why This Is Cool

Most perturbation models answer one cell type’s question. X-Cell answers across the disease-relevant landscape. The power-law scaling finding matters: it means this approach will keep improving with more data and compute - the same engine that drove LLM breakthroughs.

📖 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:

BioHackathon Edinburgh 2026 | March 20-22, Edinburgh

Three days at the University of Edinburgh bringing together life scientists, programmers, and industry partners to hack on real biological challenges. Tracks cover academic research (gene regulation, drug discovery, imaging), industry innovation, and a non-coder track for experimental design and project management. Applications are closed, but one to watch if you’re at a UK university for next year.

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 going deep on a corner of drug discovery that rarely gets attention outside specialist circles: the computational tools being built specifically for peptides.

I came across Rodrigo’s review paper while researching the next primer in our series, and what struck me was how much progress has happened quietly in this field: new notation systems, fingerprinting methods, open-source toolkits, largely invisible to anyone not working directly in it. With peptide therapeutics drawing serious investment right now, it felt like exactly the right moment to bring it to a broader audience.

We got on a call, and the rest became this primer.

What’s your biggest time sink in early drug discovery process?

Get open-source tools

Peptides aren’t small molecules. They aren’t proteins. And for a long time, the computational tools didn’t know what to do with them.

For this primer, we spoke with Rodrigo Ochoa, a Senior Scientist at Novo Nordisk working on peptide design using AI, cheminformatics, and structure-based methodologies, and lead author of a new review paper on the open-source tools now available for peptide analysis.

Semaglutide did not become a blockbuster by accident. The GLP-1 receptor agonist behind Ozempic and Wegovy is a peptide - but not a natural one. Its half-life was extended from minutes to days through deliberate chemical modifications: a fatty acid chain, non-natural amino acid substitutions, structural tweaks that made it resistant to enzymatic degradation. The result was a molecule that could be dosed once a week and reach hundreds of millions of patients.

Designing the next one requires computational tools that barely existed a decade ago. A field called peptide cheminformatics is building them.

🧬 Why Peptides Need Their Own Computational Tools

The pharmaceutical industry has mature computational tools for small molecules. Biologics - proteins and antibodies - have their own tooling built around sequence analysis and structural biology. Peptides sit in between, and that middle ground has historically been underserved.

The scale of the problem compounds this. Medicinal chemists routinely work with non-canonical amino acids - hundreds of modified or synthetic building blocks beyond the standard 20 - that can dramatically alter a peptide’s properties. The resulting chemical space is so vast it cannot be explored broadly. In practice, scientists work within local regions: making targeted modifications around a known scaffold. That makes the quality of computational tools critical. Every prediction matters more when experiments are limited.

🧩 How Peptides Are Written Down, And Why It Matters

Before you can predict anything about a molecule, you need to represent it. For peptides, this has been the most fundamental bottleneck.

FASTA notation handles natural amino acids but cannot encode chemical modifications. SMILES can represent any chemical structure but loses the underlying sequence logic. Neither captures what peptide chemists actually need.

“The main issue is that we need a mix between a sequence-based representation and a full-atom representation,” Ochoa says. “Special notations like HELM or BILN rely on what we call a monomer dictionary - it contains the chemical information of each monomer and how each monomer can be connected to each other. That way we can make a full-atom representation of a whole peptide in a systematic way.”

HELM notation, developed by the Pistoia Alliance in 2012, was the first major standardisation milestone. BILN, published in 2023 from work at Boehringer Ingelheim, took a more practitioner-friendly approach. “What we tried to do was make it as friendly as possible for the medicinal chemist,” Ochoa says - “something complex enough to include modified molecules, but that the chemist working with it can understand.” A Python package called pyPept now enables interconversion between notations, which matters when data arrives from multiple sources with different conventions.

🧮 From Fingerprints to Embeddings

Once you have a representation, you need to measure similarity - to ask whether a proposed modification is moving you toward or away from a desired property.

The standard approach for small molecules is Morgan fingerprints: fast, interpretable, and well-validated. But “you can lose some context about analogues or natural variants of the molecule,” Ochoa notes. The detection radius that works for small molecules doesn’t scale to the more complex architecture of peptides. Alternatives like MAP4 and monomer-based fingerprints address this by encoding similarity at the level of peptide building blocks rather than individual atoms.

The more significant shift is toward embeddings. Protein language models like Meta’s ESM2 capture evolutionary and contextual information rather than explicitly encoded chemical structure.

In practice the two approaches complement each other. Fingerprints work well for simpler tasks like solubility prediction. Embeddings outperform on complex ones like binding affinity, where the relevant signal is harder to specify in advance.

⚙️ Peptide Cheminformatics Tools in Drug Discovery

Matched molecular pair analysis identifies which specific modifications statistically improve a given property. “The methodology lets you see how a small change in the chemical structure can have a big positive or negative impact on a potency readout or a physicochemical property,” Ochoa says. Clustering algorithms help design diverse compound libraries. Together, these tools support lead optimisation workflows from hit identification to refined candidates.

PepFuNN, an open-source toolkit from Novo Nordisk built on RDKit and biopython, is a direct example of industry contributing back to the community. “It’s a way for industry to contribute to the academic field,” Ochoa notes, “because at the end we also use the open-source methodologies available in the state of the art.” Web servers exist for researchers without programming skills - the value of these tools extends beyond computational specialists.

📊 Why Peptide Data Remains the Biggest Barrier

The field’s most underappreciated constraint is not methods - it is data. Public peptide datasets are scattered and inconsistently annotated. Commercial datasets are more standardised but confidential. ChEMBL is actively working to annotate peptide entries with HELM notation, but it remains ongoing work. Even excellent models are only as good as what they are trained on.

🔮 Generative AI and the Future of Peptide Design

Generative models are arriving. PepINVENT uses reinforcement learning to propose synthesisable peptides optimised for specific properties. The hallucination problem is real - models can generate chemically unrealistic structures - and knowing when a model is operating within its reliable range is critical.

The near-term priorities: embeddings that properly capture non-canonical chemistry, generative models trained on better data, and LLM-based agents that lower the barrier for medicinal chemists without deep computational expertise. The direction is clear - from tools for specialists to tools for everyone who designs peptides.

Peptide therapeutics are drawing serious investment. Semaglutide is the headline, but behind it is a broader pipeline - GLP-1 analogues, cyclic peptides, stapled peptides, peptide-drug conjugates. The computational infrastructure is catching up. The data problem remains. But the foundations are solid enough to build on. Rodrigo’s paper is one of the first systematic attempts to map these tools in one place, and it arrives at exactly the right moment.

💬 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

ExplainBind: Interpretable protein-ligand binding prediction grounded in non-covalent interactions

🔬 Protein-ligand binding sits at the heart of drug discovery. Most ML models treat it as a pattern-matching problem, offering predictions without chemical insight. You get a score, not a mechanism.

ExplainBind changes that. Developed by researchers at the University of Glasgow, the Broad Institute, and Harvard, it predicts binding by explicitly modelling the non-covalent interactions (hydrogen bonds, hydrophobic contacts, electrostatic forces, van der Waals) that drive molecular recognition.

🧬 The model supervises token-level cross-attention using non-covalent interaction maps derived from PDB complexes, linking predictions to physicochemical forces at play. It operates from protein sequences and ligand representations, with no 3D structure required.

⚡ Evaluated on Angiotensin-Converting Enzyme and L-2-hydroxyglutarate dehydrogenase, ExplainBind achieved an AUROC of 0.993, outperforming eight baseline models, and enriched potent ACE inhibitors at the top of virtual screening rankings.

🎯 Applications & Insights

1️⃣ Interpretable Binding Prediction

AUROC of 0.993 against eight baselines, grounded in physicochemical interaction modelling rather than statistical correlation. Applies to proteins without solved 3D structures.

2️⃣ Binding Pocket Localisation from Sequence

Infers binding pockets from sequence data alone by analysing attention patterns between residues and ligand atoms. No crystal structure required.

3️⃣ Retrospective Ligand Prioritisation

ACE inhibitors enriched at the top of virtual screening rankings, demonstrating practical utility for filtering large chemical libraries.

4️⃣ Residue-Level Mechanistic Insight

Attention mechanisms map amino acid residues to ligand atoms in patterns corresponding to real non-covalent interactions, revealing which residues drive binding.

💡 Why This Is Cool

Interpretability in drug discovery usually means post-hoc explanations bolted onto black-box models. ExplainBind builds the chemistry in from the start. Supervising attention with non-covalent interaction maps is a cleaner approach, and the residue-level outputs are useful for medicinal chemists, not just reviewers.

📖 Read the paper

💻 Try the code

MACE-POLAR-1: A polarisable electrostatic foundation model for molecular chemistry

🔬 Short-range ML potentials miss the long-range electrostatic effects governing solvation, protein-ligand binding, and transition metal chemistry. Standard MLIPs treat charges as fixed, breaking down wherever electron density redistribution matters.

MACE-POLAR-1 extends the MACE architecture with explicit polarisation and charge equilibration, trained on OMol25: 100M hybrid DFT calculations.

🧬 A non-SCF induction formalism and learnable Fukui functions handle charge/spin equilibration, decomposing energy into local, non-local, and electrostatic terms across variable charges, spins, and ions.

⚡ Results: fourfold improvement on protein-ligand interactions (PLF547, MAE 0.37 kcal/mol), sub-kcal/mol crystal lattice energies, and accurate liquid water density from 270-330 K.

⚗️ Applications & Insights

1️⃣ Protein-Ligand Binding

Fourfold improvement over short-range baselines on PLF547, capturing long-range electrostatics critical to binding affinity.

2️⃣ Solvation and Transition Metal Chemistry

Accurate solvation structure of Fe ions, redox chemistry, and ionisation energies of hydrated transition metals.

3️⃣ Molecular Crystals

Sub-kcal/mol lattice energy errors on X23-DMC, where long-range electrostatics govern crystal packing.

4️⃣ Electric Field Response

Correctly models applied electric fields, enabling electrochemical and polarisation-sensitive applications.

💡 Why This Is Cool

Foundation models for chemistry have largely ignored polarisation. MACE-POLAR-1 puts it at the centre, delivering chemical accuracy across thermochemistry, condensed matter, and biomolecular systems.

📖 Read the paper

💻 Try the code

🤖 Get the model

Bringing the Genetically Minimal Cell to Life on a Computer in 4D

🔬 JCVI-syn3A, the genetically minimal bacterium with just 493 genes, is the simplest living organism. Researchers at UIUC, Johns Hopkins, Harvard, and the J. Craig Venter Institute built the first 4D whole-cell model (4DWCM) of its ~100-minute cell cycle.

🧬 Three simulation engines run coupled: RDME for stochastic gene expression, CME/FBA for metabolic kinetics, and Brownian dynamics for chromosome replication and segregation.

⚡ 50 replicate simulations recover the ~105-minute doubling time, ribosome counts, mRNA distributions, and spatial heterogeneity, validated against sequencing, proteomics, and imaging. Each cell is unique.

🔬 Applications & Insights

1️⃣ First Complete 4D Cell Cycle Simulation

All 493 genes, full metabolism, chromosome replication, ribosome biogenesis, and division modelled in space and time.

2️⃣ Hybrid Multi-Scale Framework

RDME, CME/FBA, and Brownian dynamics fully coupled: a blueprint for whole-cell computational biology.

3️⃣ Stochastic Heterogeneity Across Replicates

50 simulated cells show unique molecular partitioning at division, predicting variability bulk experiments cannot resolve.

4️⃣ Multi-Omics Validated

Ribosome counts, mRNA half-lives, protein distributions, and division timing match sequencing, proteomics, imaging, and cryo-ET data.

💡 Why This Is Cool

Whole-cell modelling has been an aspiration for decades. This is the closest any model has come to a complete, spatially resolved simulation of life.

📖 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.

Recap of Berlin Bio AI Hackathon (Feb 27-28) with Hannah Payette Peterson, one of the organisers:

Last weekend, 76 researchers from across Europe gathered in Berlin for a 24-hour biotech x AI hackathon co-organised by Nucleate Germany, JUNI, and Project Europe. Three challenge tracks ran in parallel: Protein Design with Adaptyv Bio 🧪, Genome Modelling and Synthesis with Serova 🔬, and Agentic AI in Life Sciences with Google DeepMind 💡

The overall winner, Omnomnomics, was a team that formed on day one from five strangers across different countries 🌍 The Agentic AI track winner, Delta Wave x NanoSpec, built an agent to classify longevity research evidence using Gemini and DeepMind’s MedGema ⚡ A Stockholm edition is already being planned for later this year, watch out for more details on this. 🇸🇪

More upcoming events:

BioHackathon Edinburgh 2026 | March 20-22, Edinburgh

Three days at the University of Edinburgh bringing together life scientists, programmers, and industry partners to hack on real biological challenges. Tracks cover academic research (gene regulation, drug discovery, imaging), industry innovation, and a non-coder track for experimental design and project management. Applications are closed, but one to watch if you’re at a UK university for next year.

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

What’s your biggest time sink in early drug discovery process?

Get open-source tools

Drug discovery has a productivity problem, and it is not the one people usually talk about.

The bottleneck is rarely a shortage of data or computational tools. It is the gap between them. You have a model that predicts toxicity, another that generates novel structures, a third that searches the literature. None of them talk to each other. Every handoff between tools is a handoff between people, and people are slow.

We spoke to Srijit Seal, visiting scientist at the Broad Institute of MIT and Harvard and lead author of a new paper on agentic AI in drug discovery, about what these systems can actually do today, where pharma is in adopting them, and why he thinks the next decade looks very different from the last.

“As LLMs get better at reasoning, can they determine which tools should be used for what kind of tasks? That’s the core question we set out to explore.”

— Srijit Seal

🤖 What Makes AI “Agentic”?

Before getting into applications, it is worth pinning down what the word means, because it gets thrown around loosely. Seal’s working definition is practical: an agentic system has access to tools and makes decisions about which ones to use. It executes functions. It does not just produce text.

“They have to have access to tools and make decisions on which tools to use,” he explains. “Not just summarise the next token and write down text. We need them to constantly use some tools to inform what they want to write next.”

That distinction matters quite a lot. A language model that summarises a paper is useful. An agent that retrieves the paper, extracts the relevant assay data, cross-references it against a compound database, runs a toxicity prediction, and writes the summary is a different category of thing. The first is a tool. The second is closer to a junior scientist.

Seal and his co-authors identify four tool types that matter in drug discovery: perception (querying databases, retrieving literature), computation (running predictions, simulations, docking), action (controlling lab hardware), and memory (storing findings across sessions so the agent builds on what it has already learned). Together they enable a closed loop: gather evidence, compute predictions, run experiments, remember results, and repeat.

🏗️ The Architectures Worth Knowing

How you wire agents together shapes what they can do. The simplest design is ReAct, short for Reasoning and Acting: the model reasons, calls a tool, observes the result, reasons again. Fast and flexible, good for focused tasks like literature triage or compound queries. Its weakness is coherence on tasks that sprawl across many steps.

Reflection architectures run multiple models in dialogue, each critiquing the other. Slower and more compute-intensive, but meaningfully better for tasks that need strategic rigour, like mapping out a multi-step synthetic route. Supervisor architectures model a research team: one orchestrating agent breaks a complex task into pieces and delegates to specialist sub-agents, each with its own domain. Swarm architectures remove the central coordinator entirely, so agents communicate peer-to-peer. That scales well across institutions, though coordination overhead grows with it.

For most of the real-world applications Seal describes, supervisor architectures do the heavy lifting. Drug discovery spans biology, chemistry, clinical data, and business context. No single agent holds all of it.

⚙️ What Agentic AI Is Already Doing

Seal’s paper presents case studies from seven startups, including ourselves, working across the drug discovery pipeline.

In literature and structure-activity relationship analysis, a supervisor-based system compressed weeks of manual review into hours, retrieving kinase selectivity data, identifying assay inconsistencies, and producing structured, cited reports.

In toxicity assessment, a ReAct-based agent combined predictive toxicology models with literature retrieval to evaluate endocrine disruption risk, reconciling computational findings with regulatory evidence in a fraction of the usual time.

In lab automation, an agent designed an automated qPCR assay for AAV quantification. Within two hours it generated a curated literature review, a MIQE-aligned protocol, and executable robot code, a process that typically takes one to four months.

More ambitiously, integrated platforms are beginning to run multi-step preclinical workflows: disease landscape analysis, target prioritisation from public omics data, structural modelling, and ranked hit generation, all in hours rather than weeks.

The common thread is not intelligence in isolation, but orchestration across tools.

⚠️ Why This Is Still Hard

None of this means the problem is solved. Seal is direct about where the friction is.

Drug discovery data is unusually difficult to reason over. Unlike images or text, it is deeply conditional: the measured activity of a compound depends on the assay format, the concentration, the cell line, the lab running the experiment. The same molecule can produce IC50 values that differ by an order of magnitude across the literature, not because the measurements are wrong, but because they are measuring slightly different things under different conditions. Getting an agent to reason reliably across this kind of data requires careful engineering, not just a capable model.

Security is a genuine concern. Agentic systems that browse external sources are vulnerable to prompt injection: malicious instructions embedded in a document or webpage that cause the agent to take unintended actions. When your agent has access to proprietary compound libraries and can exfiltrate data externally, the attack surface is not theoretical. Hallucinations, while less frequent in modern models, remain a risk in high-complexity scenarios, exactly the kind where agentic AI is most useful. And there is no mature benchmarking framework for agentic drug discovery systems yet, which makes it hard to compare capability claims rigorously.

💊 Where Pharma Is

The gap between what these systems can demonstrate and what is running inside large pharmaceutical companies is wide. That is not a criticism; it reflects appropriate caution in an industry where errors have consequences for patients.

Seal sees a clear pattern: initial scepticism from scientists and programmers, rapid conversion once they try the tools on real data, slow movement through institutional sign-off. “Once they try it on dummy data, they are like: this is actually quite good. I can automate a lot of my work.” The bottleneck is not conviction. It is process.

Traction right now is concentrated in lower-stakes tasks: literature summarisation, report drafting, document search. One underappreciated effect is democratisation: agentic tools give a small biotech with one computational scientist access to capabilities that previously required a whole team. “Previously you would have one person telling me to dock on AlphaFold, another to predict toxicity, another to automate the robot,” Seal says. “It is quite difficult to find someone with all those skills.”

The companies investing most visibly, AstraZeneca, Eli Lilly, GSK, are largely building internal infrastructure and running pilots. Industry watches academic and startup work carefully and pulls selectively. “Some companies will wait to see which one Lilly chose. If a bigger company signs a deal, the others follow.”

🔮What Comes Next

The near-term direction is self-driving laboratories: closed-loop systems that design experiments, instruct hardware, analyse results, and update hypotheses without human sign-off at each step. Prototypes exist. Scaling them to multi-step workflows with the reliability pharma demands is the engineering challenge.

New infrastructure is arriving to support agent collaboration across organisations. Anthropic’s Model Context Protocol (MCP) and Google’s Agent2Agent (A2A) protocol, now consolidated under the Linux Foundation, are building standardised interfaces so agents from different companies and labs can exchange data regardless of their underlying stack. The longer-term vision is discovery pipelines where a target identification agent at one institution hands off to a synthesis agent at another without a human in the middle.

Seal’s most striking prediction is what he calls the agentic portfolio. A scientist’s accumulated knowledge, preferred protocols, learned intuitions about which assay conditions are reliable, encoded as instructions that travel with them between jobs. When you hire someone, you hire their agents too. It sounds futuristic until you consider that this already happens informally whenever a scientist moves labs and brings their scripts, contacts, and institutional memory. The agents just make it explicit.

The consistent theme across all of it: these tools amplify human judgement, they do not replace it. The scientist who knows why a synthesis route is problematic, who recognises when a model’s training data does not reflect the patient population, who understands the difference between a statistically significant and a clinically meaningful result: that expertise remains the rate-limiting step. Agentic AI handles the orchestration. The thinking that matters most is still ours.

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

What’s your biggest time sink in the drug discovery process?

Get open-source tools

ChemAudit: Open-source chemical structure validation for chemistry ML pipelines

🔬 Bad chemistry data doesn’t announce itself. Incorrect stereochemistry, undefined stereocenters, PAINS-flagged compounds, inconsistent tautomers. These issues don’t throw errors. They just quietly degrade your models, and you find out downstream when it’s already too late.

ChemAudit is a free, open-source web platform consolidating structure validation, standardisation, structural alert screening, and ML-readiness scoring in one interface. No command line required.

🧬 Built on RDKit, MolVS, and the ChEMBL structure pipeline, it runs 15+ validation checks, screens 480+ PAINS patterns and 700+ pharmaceutical filters, and scores ML-readiness from 0-100 across 451 molecular descriptors and 7 fingerprint types. Batch processes up to 1M molecules. MIT-licensed and self-hosted.

⚗️ Applications & Insights

1️⃣ End-to-End Structure Validation

15+ checks across parsability, valence, stereochemistry, and representation consistency catch errors before they propagate into training data or downstream analyses.

2️⃣ Structural Alert Screening at Scale

480+ PAINS patterns and 700+ pharmaceutical filter rules flag problematic compounds, traceable to BMS, Glaxo, Dundee, and ChEMBL, across drug discovery and natural products workflows.

3️⃣ Quantitative ML-Readiness Scoring

Scoring against 451 descriptors and 7 fingerprint types gives an objective measure of dataset quality. Not just structural validity, but actual usability in ML pipelines.

4️⃣ Drug-Likeness and ADMET in One Place

Lipinski, QED, Veber, Ghose, and Muegge alongside ADMET predictions for synthetic accessibility, solubility, and CNS penetration. No switching between tools.

💡 Why This Is Cool

Anyone who has cleaned a chemical dataset knows the fragmented toolchain this replaces. ChemAudit consolidates the entire workflow into one accessible interface backed by the most trusted open-source chemistry libraries. Built for scale and free to use.

📖 Try the tool

💻 Try the code

TrialBench: Multi-modal AI-ready datasets for clinical trial outcome prediction

🔬 Over 90% of clinical trials fail. The data that could predict those failures (dropout rates, adverse event patterns, dosing miscalculations, duration overruns) already exists. It’s just fragmented, inconsistently formatted, and nearly impossible for ML researchers to use.

TrialBench changes that. Published in Nature Scientific Data, this benchmark delivers 23 curated, multi-modal, AI-ready datasets spanning eight clinical trial prediction tasks, assembled from ClinicalTrials.gov, DrugBank, TrialTrove, and ICD-10.

🧬 The eight tasks span the full trial lifecycle: duration, dropout, adverse events, mortality, trial approval, failure reasons, eligibility criteria design, and drug dose finding. Each dataset pairs structured metadata, molecular drug representations, and free-text clinical descriptions, the same mix a human reviewer works with.

⚡ Open Python and R packages handle loading and benchmarking, with baseline models and standardised splits for every task, enabling direct comparison without rebuilding evaluation pipelines.

📊 Applications & Insights

1️⃣ Standardised Benchmarking Across the Trial Lifecycle

Eight tasks covering every major failure point, from protocol design to post-trial outcomes, in one resource. No bespoke dataset assembly required.

2️⃣ Multi-Modal Data Integration

Structured metadata, molecular representations, and free-text clinical descriptions come packaged together: the heterogeneous mix real-world prediction actually demands.

3️⃣ Failure Prediction at Multiple Levels

Tasks span early failure modes (dosing, eligibility design) and late-stage outcomes (adverse events, mortality, approval), a framework for understanding where trials break down.

4️⃣ Reproducible Baselines for Fair Comparison

Standardised splits and baseline implementations let new methods be directly compared. No need to rebuild evaluation pipelines from scratch.

💡 Why This Is Cool

Clinical trial prediction has been a graveyard for AI benchmarks: data siloed, endpoints inconsistent, results near-impossible to reproduce. TrialBench addresses all three at once. It won’t solve clinical AI alone, but it gives the field a common language.

📖 Read the paper

💻 Try the code

STRIPES: A symbolic language of molecular recognition for structure-informed drug design

🔬 Protein-ligand binding isn’t a static snapshot. It’s a dynamic interplay of interactions that shifts over time. Most drug design tools treat it as one. The information locked inside MD trajectories is largely untapped.

Researchers at Eindhoven University of Technology developed STRIPES, encoding protein-ligand interaction dynamics from MD simulations as symbolic token sequences, treating molecular recognition as a language for retrieval, comparison, and generation.

🧬 STRIPES converts interaction fingerprints: hydrogen bonds, hydrophobic contacts, pi-stacking, electrostatics, into compact symbolic strings over time. These form a physically interpretable record of binding, enabling similarity retrieval and conditioning of de novo generation on target interaction profiles.

⚡ Validated across JAK1, PIM1, PPARd, and the androgen receptor on ~15,000 MD trajectories: STRIPES-guided Transformer design produced confirmed hits at picomolar to nanomolar potency. Conditioning on interaction patterns opens routes conventional methods miss.

🔬 Applications & Insights

1️⃣ Interaction-Conditioned De Novo Generation

STRIPES enables ligand generation conditioned on specific interaction profiles, moving beyond shape and pharmacophore to dynamic binding behaviour captured from real MD trajectories.

2️⃣ Physically Interpretable Representations

Each token maps to a real interaction type. Unlike black-box embeddings, STRIPES remains interpretable, supporting hypothesis-driven design and mechanistic insight into binding.

3️⃣ Trajectory-Level Similarity Search

STRIPES sequences enable fast comparison of binding dynamics across large MD datasets, identifying compounds with similar interaction fingerprints regardless of structural similarity.

4️⃣ Experimental Validation Across Four Targets

Compounds were synthesised and tested on JAK1, PIM1, PPARd, and AR, confirming picomolar to nanomolar activity and going beyond benchmarking to real biological validation.

💡 Why This Is Cool

STRIPES makes MD trajectories a generative resource, not just an analytical one. Encoding binding dynamics as a learnable language adds a physically grounded layer to drug design that static docking can’t offer.

📖 Read the paper

💻 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.

Last week we featured two Adaptyv Bio hackathons - check back soon for updates on results.

BioHackathon Edinburgh 2026 | March 20-22, Edinburgh

Three days at the University of Edinburgh bringing together life scientists, programmers, and industry partners to hack on real biological challenges. Tracks cover academic research (gene regulation, drug discovery, imaging), industry innovation, and a non-coder track for experimental design and project management. Applications are closed, but one to watch if you’re at a UK university for next year.

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

What’s your biggest time sink in the drug discovery process?

Get open-source tools

MULTI-evolve: Accelerating protein engineering through predictive epistasis

🔬 Protein engineering traditionally requires multiple rounds of directed evolution to identify beneficial mutations, a time-consuming and resource-intensive process.

Arc Institute’s MULTI-evolve bridges predictive biological modelling with functional laboratory optimisation to accelerate protein engineering. By integrating computational mutation prediction with targeted experimental validation, the framework identifies and combines beneficial mutations without requiring repeated rounds of directed evolution.

🧬 MULTI-evolve addresses epistasis - where mutation effects depend on one another - by experimentally testing pairwise mutation combinations and training models to predict higher-order interactions. This enables efficient design of synergistic multi-mutant variants with substantially enhanced activity.

⚡ The framework was applied to APEX, CRISPR-Cas13d, and an anti-CD122 antibody. MULTI-evolve first identified improved single mutants (up to threefold improvement) and leveraged epistatic modelling to design variants containing up to seven mutations, achieving improvements of up to 256-fold.

Applications & Insights

1️⃣ Improving Discovery of Function-Enhancing Mutations Combining protein language models with targeted validation, MULTI-evolve prioritises beneficial single mutations before combinatorial optimisation begins.

2️⃣ Modelling of Epistasis Quantifying epistatic interactions via pairwise mutation testing enables prediction of synergistic multi-mutant variants without extensive screening.

3️⃣ Single Round Design of High-Order Mutants Single-cycle generation of multi-mutant variants accelerates development, reduces laboratory workload, and eliminates large combinatorial libraries.

4️⃣ Multi-Mutant Engineering of Diverse Proteins MULTI-evolve has enhanced APEX via a valine substitution improving hydrophobicity and sodium ion coordination, engineered a dCas-Rx splicing tool with improved RNA binding and crRNA recognition, and demonstrated promise in multi-objective antibody design with improved binding and expression.

💡 Why This Is Cool

MULTI-evolve unites predictive modelling with targeted experimental validation to streamline protein engineering. By quantifying epistatic interactions, it enables single-cycle optimisation, reduces laboratory burden, and demonstrates broad utility across enzymes, CRISPR tools, and antibodies.

📖 Read the paper

💻 Try the code

Deep Ignorance: Filtering dangerous knowledge at training time

🔒 What if the safest AI model is simply one that never learned the dangerous stuff in the first place? That’s the question researchers at EleutherAI, the UK AI Security Institute, and Oxford’s OATML group decided to take seriously and honestly, the results are pretty compelling.

Here’s the thing: open-weight models can be downloaded and fine-tuned by anyone. Every post-training safety fix tried so far gets undone within a few hundred steps. The weights remember what they saw during training, and that memory is surprisingly easy to recover.

🧪 Applications and Insights

1️⃣ Tamper-resistant biothreat suppression Filtered 6.9B models held up against 10,000 fine-tuning steps and 300M tokens of adversarial biothreat text, more than ten times the resistance of any post-training baseline ever tested.

2️⃣ No capability trade-off Filtering costs less than 1% of total training FLOPs. Performance on MMLU, HellaSwag, PIQA, and LAMBADA? Basically flat.

3️⃣ Filtering and Circuit Breaking are better together Filtering alone can’t stop in-context retrieval attacks. Pair it with Circuit Breaking and that gap closes. Neither works alone, but together they’re genuinely robust.

4️⃣ It doesn’t work for everything Filtering behavioural propensities like toxicity and jailbreak compliance is a different beast. Filtered models can actually get more vulnerable to few-shot attacks on those tasks. Knowledge and behaviour are not the same problem.

💡 Why This Is Cool

The field has spent years trying to unlearn dangerous knowledge after the fact. This paper just asks: why teach it at all? Simple idea, rigorous execution, and refreshingly honest about where it falls short.

📖 Read the paper

💻 Try the code

iSight: Automated IHC staining assessment at scale

🔬 Stanford Healthcare runs over 3,500 immunohistochemistry cases every single month. Back in 2005, that number was 700. The workload has quintupled and the workforce hasn’t kept up. You know what that means in practice: more pressure, more variability, more room for error.

Researchers at the University of Pennsylvania and Stanford built iSight to help. It’s a multi-task AI model for automated IHC staining assessment, trained on HPA10M, a freshly curated dataset of over 10 million IHC images across 45 normal tissue types and 20 cancer types. Nothing like it has existed at this scale before.

🧫 Applications and Insights

1️⃣ Three staining tasks, one model iSight predicts intensity, localisation, and quantity simultaneously, hitting 85.5% accuracy for location, 76.6% for intensity, and 75.7% for quantity, beating fine-tuned PLIP and CONCH by 2.5–10.2%.

2️⃣ Calibration you can actually trust Expected calibration errors sat between 0.015 and 0.041. The confidence scores reflect real accuracy, which matters enormously when a clinician is deciding whether to act on a prediction.

3️⃣ It makes pathologists more consistent Eight pathologists evaluated 200 images before and after seeing iSight’s suggestions. Inter-rater agreement jumped from Cohen’s κ 0.63 to 0.70. The AI didn’t override anyone, it just quietly anchored the room.

4️⃣ Holds up under messy real-world conditions Salt-and-pepper noise, tissue fold artefacts, scan dropouts, tested across four severity levels, performance never deviated more than 1.5% from baseline.

💡 Why This Is Cool

Most pathology AI papers benchmark and stop there. This one actually put the model in front of clinicians and watched what happened. The finding that pathologists improve but still trail the AI alone is genuinely fascinating, it says as much about human decision-making as it does about the model.

📖 Read the paper

💻 Try the code

🗓️ Events & Competitions

We’re launching a new section of the newsletter dedicated to tracking the best competitions, hackathons, and community challenges happening!

If you’re looking to get hands-on with cutting-edge targets, and have your work experimentally validated, these are the events worth your time. We’ll be featuring these regularly, so if you know of something worth highlighting, reply and let us know!

First up, two hackathons happening this weekend courtesy of our friends at Adaptyv Bio 👇

1. bioArena x Adaptyv: Agents for Protein Design Hackathon | Feb 28, San Francisco

   Teams are using AI agents to design binders against TREM2, one of the strongest genetic risk factors for Alzheimer’s disease and a highly active area of drug discovery. Top designs get experimentally validated by Adaptyv in the wet lab.

2. Berlin Bio x AI Hackathon | Feb 27-28, Berlin

   24 hours, an aging-relevant target, and $10,000 in Adaptyv lab credits on the line. The top 100 designs go straight to Adaptyv’s wet lab for experimental validation, with additional tracks on genome modeling and agentic AI in life sciences. Applications are closed but we’ll be watching the results closely.

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

What’s your biggest time sink in early drug discovery process?

Get open-source tools

Artificial intelligence is sweeping through healthcare. Investment is rising, publications are being released daily, and new tools are starting to enter real clinical settings. As the pace accelerates, one question keeps resurfacing: what’s the role of clinicians in all of this? And more importantly, how involved do they need to be for AI to actually improve care?

This piece, by medical student Maria Aquilue Dies, looks at the gap between clinicians and AI development, how current tools are being adopted, and what needs to happen next to build a truly collaborative, clinically grounded AI ecosystem.

📚 The Knowledge Gap

Clinicians and AI developers do collaborate. Large hospitals with research units host multidisciplinary teams, and medical AI companies often call on clinicians as consultants. However, significant differences between the clinical and technological domains persist, and they often continue to operate alongside one another rather than in close integration.

When clinicians participate, their role tends to be more advisory than hands-on, in part because many of the underlying concepts (machine learning fundamentals, deep learning architectures, neural networks, transformers) remain unfamiliar territory. Without a solid understanding of how these systems work, it’s harder to contribute meaningfully to design decisions, evaluation criteria, or safe deployment strategies.

This gap has some meaningful consequences:

• Promising applications remain unrealised because clinicians, the people closest to the problems, are not always fully aware of the extent of what AI can offer.

• Tools under development progress without vital clinical insights, leading to solutions that miss workflow realities or fail to account for common pitfalls.

• Validated tools struggle to reach patients because many clinicians aren’t aware they exist, and therefore may not request them, or don’t feel confident using them.

• And importantly: clinicians are ultimately responsible for the ethical use of AI in patient care. Without literacy in how these systems behave, informed oversight becomes impossible.

To illustrate these points, consider the daily burden of “data dredging” through fragmented medical records to synthesise a complex patient’s history. An LLM-based chatbot integrated directly into the Electronic Health Record (EHR) interface could instantly streamline this, yet many hospitals have still not implemented it simply because clinicians are unaware that such a dynamic interaction is even possible. When these tools are developed in isolation, they often fail to adapt to the specific nuances of different medical specialties or feature cumbersome, poorly integrated interfaces for the average user, leading to immediate abandonment. Furthermore, without a foundational understanding of the technology, clinicians may either reject the tool out of “black box” skepticism or, conversely, over-rely on its output without realizing the system might prioritize repetitive data over clinically significant outliers.

Yet there’s a positive side to all this. Despite limited formal training, clinicians are generally optimistic about AI. Many want to try new tools and collaborate with developers. They simply lack the pathways, education, and institutional structures to do it.

This is why medical training needs to evolve. First and foremost in medical school, but also throughout residency and continuing professional development. When clinicians understand the technology shaping their field, they push for safer, more useful systems, and adoption becomes smoother and more equitable.

🧭 Illustrating the Transition

The Present Landscape

Today’s AI adoption in hospitals is uneven. Most clinicians use only a narrow set of tools, often as end‑users rather than co‑designers. But even these early steps show how AI can shift workflows in meaningful ways.

Take medical chatbots, arguably the most widely adopted tool.

Before these LLM-based tools were adopted, reviewing evidence in difficult cases was a slow and repetitive process of diving into large repositories and manually filtering papers in search of a specific nugget of information. Moreover, the result was usually poorly tailored to the specific patient’s age, comorbidities, or context.

Now clinicians can ask a question with all relevant context and receive a focused, case‑specific answer grounded in the literature. The process is not only faster in a setting where acting quickly can save lives, but often more individualized, drawing on diverse sources and integrating nuances that would be cumbersome to search manually. That said, these systems are not infallible: outputs still require clinical verification, institutional oversight, and a clear understanding of their limitations, particularly around hallucinations, data privacy, and regulatory accountability.

Additionally, while chatbots may be the most widespread application, a broader range of AI tools is being slowly but steadily deployed in hospitals. According to the U.S. FDA, more than 1,000 AI/ML-enabled medical devices have now received regulatory authorisation, the majority concentrated in radiology and imaging.

Some examples that stand out are:

• AI medical imaging systems: widely used in breast cancer screening and in the detection of pathologies in chest X‑rays, such as pneumonia or lung cancer. Some hospitals use AI to automatically calculate brain tumor volume from MRI scans, replacing error‑prone manual measurements.

• Risk‑stratification models, ICU monitoring systems, tools for surgery planning and supervision (e.g., back surgery planning or neurosurgery phase recognition), and automatic EHR generation from clinician–patient conversations.

Nonetheless, an important challenge remains: what gets adopted first depends less on clinical need and more on hospital culture, peer influence, and the enthusiasm of early adopters.

What a Clinically Grounded AI System Looks Like

These applications are encouraging, but far from the end goal. A truly effective AI ecosystem requires two fundamental transformations.

First, clinicians must play an active role in guiding AI development, backed by an education that emphasizes AI literacy. This is essential to address two of the most relevant challenges in AI applications: ethical adoption and creating tools that address real-world clinical needs

Second, AI must be embedded into workflows, not stacked awkwardly on top of existing systems. The future shouldn’t be a patchwork of isolated apps scattered across hospitals. It needs:

• Seamless integration with electronic health records

• Interconnected tools across healthcare facilities

• Equitable deployment and sustainable funding

This creates real-time, longitudinal support across the entire patient journey.

Once this ecosystem exists, current applications can evolve dramatically. Medical chatbots could link directly to patient records for pre-visit summaries, inpatient follow-up, and real-time guidance. AI imaging tools could feed into multidisciplinary diagnostic pipelines, track changes over time, or predict disease progression and treatment response.

These are not futuristic dreams, but natural extensions of what we already have. Clinicians taking a clear step forward and advocating for the necessary infrastructure, education, and collaborations will be a key element in this transition.

💡How to Promote Safe and Effective AI Adoption

We now know that the implementation of AI in medical settings has both incredible potential to help and a clear need for informed clinical supervision. In this context, the question becomes how the main actors in this change, clinicians and institutions (be they public or private), can actually take the necessary steps to make the transition as smooth as possible.

The first step: make AI literacy a standard part of medical training

AI-driven systems will only become more prevalent in clinical settings, so it is the responsibility of medical school administrators to update curricula to include practical, hands-on AI training and safety principles. Hospitals should also provide programs ensuring that already graduated physicians have access to this education through residency and continuing education.

Even clinicians without formal positions in these institutions can contribute by providing mentorship to students and colleagues, and by advocating for the integration of AI training into clinical sessions, hospital programs, and conferences.

I have had the opportunity to witness the reality of this situation firsthand during my medical school years. In my faculty, as it stands, the only structured exposure to AI in medicine is relegated to a couple of elective courses. While many of us have encountered the topic through clinical sessions or specific research initiatives (such as a final year project), it remains absent from the formal, core curriculum. Looking ahead, it is essential that these concepts move beyond optional seminars and become a foundational part of our training, whether by updating existing modules or, ideally, establishing dedicated courses within the standard medical program.

Time to work: create structured spaces for collaboration

Research teams and companies developing medical AI applications should build formal multidisciplinary teams and establish fast, effective feedback mechanisms to ensure smooth development and integration. Hospital coordinators and chiefs of service can further support this process by offering incentives to departments that adopt AI responsibly and safely.

Clinicians, for their part, can actively contribute by seeking opportunities to participate in AI projects and initiating conversations within their departments, sharing insights on workflow, safety, and usability.

A way to ensure safety: build clearer frameworks for evaluating clinical AI

Regulatory bodies and hospital administrations should develop clear institutional guidelines, create accessible evaluation benchmarks and standardise performance reporting in clinical research. Clinicians can then build on this foundation by learning to critically assess AI tool reports and real‑world validation, and by participating in committees that help shape how these tools are evaluated and adopted.

Inspire adoption: highlight successful real‑world use cases

Sharing insights about new tools and the development process in multidisciplinary teams is essential to encourage implementation in an equitable and consistent manner.

This can happen on both large and small scales: professional societies should consider including clear, clinician-friendly examples of successful implementation in practice in conferences and journals, while clinicians can contribute by discussing new tools and personal experiences with colleagues. A key way to stay updated on the state of the art of AI applications in medicine is through trusted newsletters and the AI-in-healthcare sections of major journals.

✨ The Future is Bright

Despite the fears sparked by early developments in AI, recent reports show that these applications can excel at assisting doctors and patients. They can reduce burdens and improve outcomes across healthcare; but only if they are thoughtfully developed, safely regulated, and responsibly deployed.

Fortunately, positive medical AI applications continue to advance at an extraordinary pace. But its true impact depends on healthcare professionals: their knowledge, their involvement, and their willingness to guide this technology responsibly.

The next generation of medical AI shouldn’t be built just for clinicians. It should be built with and by them. And if we get that right, the future of medicine won’t just be faster or more efficient. It will be smarter, safer, and far more humane.

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

What’s your biggest time sink in the drug discovery process?

Get open-source tools

🧬 Enzyme Motif Miner: Function Annotation Through Catalytic Geometry

Identifying enzyme function by matching the geometry of catalytic sites

That is the core idea behind Enzyme Motif Miner, a tool from researchers at EMBL-EBI and Leiden University Medical Center. Rather than relying on sequence similarity, which breaks down below about 40% identity, it geometrically matches 3D arrangements of catalytic residues against a curated library of 6,780 templates from the Mechanism and Catalytic Site Atlas, M-CSA.

Every match traces back to experimentally characterised enzyme mechanisms, giving you biologically grounded annotations rather than black box predictions.

🔬 Applications and Insights

1️⃣ Outperforms sequence and structure alignment
Below 27.5% amino acid identity, sequence-based models perform no better than random. Enzyme Motif Miner stays predictive by leveraging conserved catalytic geometry, beating both amino acid and 3Di Foldseek alignment for distantly related enzymes.

2️⃣ Scales to the human proteome
Applied to 20,406 AlphaFold2-predicted human proteins, it matched 42% of annotated enzymes whilst hitting only 6.4% of non-enzymes. Of matched enzymes with UniProt active site annotations, 83.7% were correctly localised to the true catalytic site.

3️⃣ Uncovers convergent evolution
By combining low CATH fold similarity with high EC number agreement, the tool flags enzymes that independently converged on the same catalytic geometry, including metallopeptidases across seven distinct CATH superfamilies sharing antiparallel helix arrangements.

4️⃣ Fast enough for proteome-scale use
On a consumer laptop with 6 cores, it scans 1,000 human protein structures in under 4 minutes, 0.23 seconds per structure, two orders of magnitude quicker than AI-based tools like DeepFri.

💡 Why It’s Cool

Most function annotation tools are either fast but coarse, such as sequence alignment, or accurate but opaque, such as deep learning. Enzyme Motif Miner sits in a useful middle ground. It is knowledge driven, interpretable, and fast enough for AlphaFold-scale datasets.

The convergent evolution detection is particularly exciting. It suggests chemistry constrains biology in ways we can now systematically measure across entire proteomes.

📖 Read the paper
💻 Try the code

🧬 RESPLICE: Programmable RNA Rewriting Without Genome Editing

Rewriting RNA without touching the genome

That is the promise behind RESPLICE, a new tool from researchers at the Arc Institute, UC Berkeley, and Stanford University.

Most RNA editing tools can knock things down or swap single bases. But what if you need to replace an entire exon. That is where trans-splicing comes in, and it has been a frustratingly hard problem for decades.

RESPLICE changes that by combining two orthogonal CRISPR effectors, one to guide the trans-splicing reaction and one to block the competing cis-splicing reaction. The result is a programmable system that can write large RNA cargoes, up to 2.1 kb, into endogenous human transcripts with real efficiency.

🔬 Applications and Insights

1️⃣ High efficiency into endogenous transcripts
RESPLICE achieved up to 47% trans-splicing efficiency in unsorted cells across 11 endogenous human transcripts in 3 cell types, and up to 90% in cells sorted for high effector expression. Prior RNA-only approaches typically topped out around 5%.

2️⃣ A second CRISPR effector supercharges efficiency
Adding the cis-splicing interfering module, CIM, boosted trans-splicing by up to 28.8-fold for some targets, TFRC from 1.6% to 47.4%. It works orthogonally to the trans-splicing module, so combining the two is additive.

3️⃣ Therapeutically relevant targets demonstrated
Proof-of-concept correction in three disease contexts, hereditary haemochromatosis about 8%, C9orf72 repeat expansion linked to ALS and FTD about 10%, and TDP43 mutation replacement about 20%. These are precisely the genes where mutational heterogeneity makes site-specific editing impractical.

4️⃣ Off-target trans-splicing is low and stochastic
Transcriptome-wide RNA-seq showed on-target specificity averaging 95.3%. Off-target events occurred at less than 0.005% of all splice junctions, well below the basal splicing error rate of more than 0.1% per intron.

💡 Why It’s Cool

RNA medicine is having a moment, but most tools either edit single bases or deliver entire transgenes. RESPLICE fills an awkward gap. It enables large, programmable, transient edits at the transcript level without permanently altering the genome.

The transience is a feature, not a bug, particularly for diseases where reversibility or tunability matters.

📖 Read the paper
💻 Try the code

🧠 MEDEA: An Agent That Verifies Its Own Omics Reasoning

Turning an omics dataset into a therapeutic hypothesis is a long, messy process

That is precisely what MEDEA is designed to fix.

Built by researchers at Harvard Medical School, the Broad Institute, and Imperial College London, MEDEA is an AI agent that takes a biological objective and executes a transparent, multi-step omics analysis using 20 specialised tools, with verification built into every stage.

Most AI agents either hallucinate intermediate steps or follow rigid templates that break down across biological contexts. MEDEA is different. It verifies context, checks tool compatibility before and after execution, screens literature for relevance, and reconciles conflicting evidence before committing to a conclusion. When evidence is insufficient, it abstains rather than guesses.

🔬 Applications and Insights

1️⃣ Cell type specific target nomination
Across 2,400 analyses spanning five diseases, rheumatoid arthritis, type 1 diabetes, Sjogren’s syndrome, hepatoblastoma, and follicular lymphoma, MEDEA outperforms standalone LLMs by up to 45.9%. It correctly resolves granular cell type distinctions that LLMs routinely collapse, such as naive versus effector memory CD4+ T cells in rheumatoid arthritis.

2️⃣ Synthetic lethality reasoning
Across 2,385 analyses in seven cancer cell lines, MEDEA improves accuracy by up to 21.7% over GPT-4o and correctly resolves over 323 cases where every tested LLM failed outright. It integrates DepMap co-dependency scores with pathway enrichment to flag gene pairs whose joint inhibition is predicted to kill cancer cells selectively.

3️⃣ Immunotherapy response prediction
Across 894 patient-level analyses from the IMvigor210 bladder cancer cohort, MEDEA achieves up to 23.9% higher accuracy than LLMs by integrating tumour transcriptomes, TMB, and microenvironment signatures. It rescues up to 50.9% of cases misclassified by the underlying ML model in the hardest subgroup, high TMB, non-inflamed tumours.

4️⃣ Verification beats volume
Ablations show a literature-only agent abstains 77.6% of the time, while an LLM-only agent abstains just 1.8% of the time but produces the most errors. The full MEDEA achieves the best accuracy with the lowest failure rate, demonstrating that structured verification beats simply adding more compute or retrieval.

💡 Why It’s Cool

MEDEA shows that the bottleneck in agentic biomedical AI is not raw reasoning power. It is the ability to stay grounded in biological context across long, multi-step analyses.

The calibrated abstention is particularly underrated. In drug discovery, a confident wrong answer is far more costly than an honest I do not know.

📖 Read the paper
💻 Try the code

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

What’s your biggest time sink in the drug discovery process?

Get open-source tools

🧪 GLDP: Latent-Space Protein Dynamics at All-Atom Resolution

🧪 Can we model protein motion without waiting hours for molecular dynamics?

That’s the goal of GLDP, a framework from researchers at EPFL and collaborators that simulates all-atom protein dynamics in latent space. Imagine teaching an AI to watch proteins move, then letting it choreograph new motions that still obey physics.

The problem with traditional MD is speed. Tracking every atom is accurate, but painfully slow, especially when rare events, like functional conformational changes, take forever to emerge. GLDP sidesteps this by simulating in a compressed representation, then decoding to full atomic detail.

🔬 Applications and Insights

1️⃣ Three propagators, one framework
They compared autoregressive neural networks, Koopman operators (linear dynamics), and score-guided Langevin methods. The neural net was the most stable for long rollouts. Langevin captured sharp side-chain detail. Koopman was lightweight, but a bit rigid.

2️⃣ Tested across scales
From alanine dipeptide to large GPCRs with 300+ residues. The autoregressive model managed 10,000-frame rollouts without collapse. Langevin recovered rotameric states with a Jensen-Shannon divergence of just 0.058.

3️⃣ Captured functional motions
On the A2A receptor, GLDP reproduced the activation pathway between inactive and active states — not just structure prediction, but full conformational switching tied to biological function.

4️⃣ Kinetic fidelity holds up
Using TICA analysis, they showed GLDP recovers physically meaningful timescales (650 steps for A1AR), where other models become unstable or lose temporal structure.

💡 Why It’s Cool

Most protein models generate snapshots. GLDP captures how proteins move, critical for drug discovery, where ligand binding often depends on timing and flexibility. By working in latent space, it’s also computationally lighter than brute-force MD, enabling rapid sampling of structural ensembles.

It’s honest about trade-offs too. Neural nets offer long-term stability. Score-based methods bring thermodynamic detail. No one-size-fits-all, and that transparency helps you choose the right tool based on your scientific goals.

📖 Read the paper
💻 Try the code

💊 CyclicMPNN: Teaching ProteinMPNN the Rules of Cyclic Peptides

Cyclic peptides are an important class of therapeutic molecules, acting as antibiotics, anticancer agents, and immunosuppressants. Compared to linear peptides, they show greater structural stability, resistance to proteolytic degradation, and improved binding specificity.

Although tools such as GenKIC and RFPeptide can generate realistic cyclic backbones, designing sequences that reliably fold into these constrained structures remains challenging.

🧬 ProteinMPNN is a powerful model for protein sequence design but was trained mainly on large, linear proteins. Consequently, it does not capture the sequence–structure relationships of small cyclic peptides, which rely on tight backbone geometry and internal hydrogen-bonding networks.

🎯 CyclicMPNN addresses this gap by fine-tuning ProteinMPNN on cyclic peptide sequence–backbone pairs validated for structural compatibility using HighFold, enabling more accurate sequence design for cyclic backbones.

🔬 Key Applications & Insights

1️⃣ Reliable Sequence Design for De Novo Cyclic Backbones
CyclicMPNN assigns sequences to GenKIC, RFPeptide, or CyclicCAE backbones in a single step, learning the rules required to satisfy tight cyclic geometry without repeated refinement.

2️⃣ Motif-Inpainting for Therapeutic Design
CyclicMPNN preserves the Nrf2 EETG motif while stabilizing the surrounding cycle, maintaining motif geometry while enabling scaffold design. This capability is critical for designing cyclic peptides that retain functional binding sites.

3️⃣ Accurate Structural Recapitulation Without Iteration
CyclicMPNN achieves ≤ 1 Å RMSD to target backbones without Rosetta relax or redesign cycles by learning true backbone compatibility. This eliminates the need for computationally expensive iterative refinement.

4️⃣ Energetically Stable Folding Funnels
Sequences designed by CyclicMPNN show strong PNear funnels, indicating genuine thermodynamic stability rather than only structural agreement. This ensures designs will fold reliably in experimental conditions.

💡 Why It’s Cool

CyclicMPNN overcomes a key bottleneck in cyclic peptide design by enabling reliable sequence assignment for constrained backbones. By learning the sequence rules unique to cyclic systems, it improves structural accuracy, stability, and motif preservation without iterative refinement.

This approach streamlines cyclic peptide design workflows and provides a framework for adapting protein design models to other constrained biomolecular systems.

📃 Read the paper
⚙️ Try the code

Big thanks to Amber Vig for writing this article, another great find!

🫀 EchoJEPA: Learning Cardiac Anatomy Without Memorising Speckle Noise

That’s the challenge EchoJEPA tackles, a foundation model from University Health Network, University of Toronto, and collaborators trained on 18 million echocardiograms. Instead of reconstructing pixels (and their noise), it predicts latent representations, learning what’s anatomically stable whilst ignoring acoustic artifacts.

Ultrasound is tricky. Speckle, depth attenuation, and acoustic shadows dominate images but tell you nothing about cardiac anatomy. EchoJEPA uses joint-embedding predictive architectures (JEPA) to downweight unpredictable noise and reinforce temporally coherent structures like chamber walls and valve motion.

🩺 Applications and Insights

1️⃣ Latent beats pixel reconstruction
In a compute-matched comparison, EchoJEPA-L outperformed VideoMAE by 27% on LVEF estimation and 45% on view classification — identical architecture and data, only the objective differed.

2️⃣ Remarkable sample efficiency
Achieved 79% view classification accuracy with just 1% of labelled data, versus 42% for the best baseline trained on 100%. A game-changer for medical AI where annotations are expensive.

3️⃣ Superior robustness
Under physics-informed perturbations (depth attenuation, acoustic shadows), EchoJEPA degraded by only 2% compared to 17% for competitors — 86% less sensitivity to artifacts that plague clinical imaging.

4️⃣ Zero-shot paediatric transfer
Trained entirely on adult data, zero-shot performance on paediatric patients (4.32 MAE) beat all baselines after fine-tuning, proving the representations genuinely capture cardiac anatomy.

💡 Why It’s Cool

EchoJEPA shows that matching your learning objective to domain physics matters enormously. By focusing on what’s predictable (anatomy) versus what isn’t (speckle), it builds representations that generalise across acquisition conditions, patient populations, and age groups. The sample efficiency with frozen backbones lowers barriers for clinical researchers lacking compute budgets for end-to-end fine-tuning.

📖 Read the paper.
💻 Try the code

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