AI SparkupAI가 과학자보다 먼저 답을 맞힌 날: 신약 개발 18개월→6개월 단축의 비결
MIT의 AI 모델 DiffDock이 100초 만에 신약 enterololin의 작용 메커니즘을 예측하고, 연구진이 6개월 만에 검증에 성공했습니다. 기존 2년/200만 달러 소요 작업을 획기적으로 단축한 AI 기반 신약 개발 사례를 소개합니다.
Elly PoretskyMy favorite molecular #protein-#ligand #docking method, #DiffDock, has been updated! The new DiffDock-L, provides a significant improvement in performance and generalization capacity.
Importantly., this new method comes with the new #DockGen benchmark, aiming to provide better evaluation metrics and help improve #generalization of #ML docking models by accounting for sequence-dissimilar proteins with very similar binding pockets in training/test splits.
Deep Confident Steps to New Pockets: Strategies for Docking Generalization
Accurate blind docking has the potential to lead to new biological breakthroughs, but for this promise to be realized, docking methods must generalize well across the proteome. Existing benchmarks, however, fail to rigorously assess generalizability. Therefore, we develop DockGen, a new benchmark based on the ligand-binding domains of proteins, and we show that existing machine learning-based docking models have very weak generalization abilities. We carefully analyze the scaling laws of ML-based docking and show that, by scaling data and model size, as well as integrating synthetic data strategies, we are able to significantly increase the generalization capacity and set new state-of-the-art performance across benchmarks. Further, we propose Confidence Bootstrapping, a new training paradigm that solely relies on the interaction between diffusion and confidence models and exploits the multi-resolution generation process of diffusion models. We demonstrate that Confidence Bootstrapping significantly improves the ability of ML-based docking methods to dock to unseen protein classes, edging closer to accurate and generalizable blind docking methods.
The Triangle AgencySpeeding up drug discovery with diffusion generative models | MIT News https://triangleagency.co.uk/speeding-up-drug-discovery-with-diffusion-generative-models-mit-news/?utm_source=dlvr.it&utm_medium=mastodon #TheTriangleAgencyNews #AlphaFold #BowenJing #DiffDock
