In this issue:
Welcome back to your weekly dose of AI news for Life Science!
This week, we have some exciting new models lined up for you:
AlphaBind: Predict and Optimise Antibody-Antigen Binding Affinity 🧬
3DSTarPred: A Web Server for Target Prediction of Bioactive Small Molecules Based on 3D Shape Similarity 💿
RapidDock: Revolutionising Drug Discovery with High-Speed Molecular Docking 💊
Dive into these game-changing innovations and explore how they are transforming the biotech and healthcare landscapes!
AlphaBind: Predict and Optimise Antibody-Antigen Binding Affinity 🧬
Synthesising antibodies is currently an intense field of research, as it represents a rapid and efficient way to fight against new viruses and diseases. However, beyond theory, the practice is more complicated, as it is challenging to optimise the best antibody candidates in a short time. In silico approaches, combining deep learning and computational biology tools, allow us to speed up the process and enable us to focus experimental efforts only on the most promising candidates, especially by efficiently screening all candidates.
Introducing AlphaBind, a software from A-Alpha Bio and NVIDIA and that utilises a protein language model to achieve state-of-the-art performance for antibody affinity optimisation!
📌 Key Insights:
AlphaBind-powered antibody optimization pipeline can deliver candidates with substantially improved binding affinity
Trained on 7.5M antibody-antigen binding measurements, including 300k unique Antibodies and 6k unique targets
Antibody optimization allowed with alternative fine-tuning Data
Efficiency of the model is demonstrated on the optimization case of Trastuzumab-scFv CDRH3 variants
3DSTarPred: A Web Server for Target Prediction of Bioactive Small Molecules Based on 3D Shape Similarity 💿
The field of drug development is, year after year, strongly influenced by in silico approaches made possible by the explosion of computational power. The goal of these software programs is to align accessibility, accuracy, and speed while maintaining simplicity for the user.
Introducing 3DSTarPred, a software recently released online and freely available, which achieves higher target prediction than the existing online software.
📌 Key Insights:
The prediction is based on 3D similarity with 1,221,364 actives from ChEMBL29 and 12,795 actives from PDBbind2020 on 5,298 proteins from ChEMBL29 and 3,016 proteins from PDBbind2020.
The algorithms include AlphaConf for molecular conformation generation and AlphaShape for 3D shape comparison and similarity calculation.
3DSTarPred is, with ADMETPred, included in AI-DrugIP, an Al-empowered platform for new drug innovation, allowing the use of 7 regression models and 20 classification models.
RapidDock: Revolutionising Drug Discovery with High-Speed Molecular Docking 💊
Accelerating molecular docking - the process of predicting how molecules bind to protein targets - could boost small-molecule drug discovery and revolutionise medicine. Unfortunately, current molecular docking tools are too slow to screen potential drugs against all relevant proteins, which often results in missed drug candidates or unexpected side effects occurring in clinical trials. Introducing RapidDock, an efficient transformer-based model for blind molecular docking.
📌 Key Insights:
Extremely fast model, processing each drug-protein interaction in just 0.04 seconds on a single GPU.
Docking ten million molecules to all human proteins on a cluster with 512 GPUs would take nine days with RapidDock, whereas DiffDock-L would take roughly 20 years, or perhaps 200 years, with a computationally demanding approach like AlphaFold-3.
RapidDock performs better than other industry-leading models like NeuralPLexer and DiffDock-L, second only to AlphaFold-3
Did you find this newsletter insightful? Share it with a colleague!
Subscribe Now to stay at the forefront of AI in Life Science.
Connect With Us
Have questions or suggestions? We'd love to hear from you!
📧 Email Us | 📲 Follow on LinkedIn | 🌐 Visit Our Website