A Perspective introducing the "Avoid-ome" — a finite set of anti-target proteins driving drug safety liabilities — and how OpenADMET combines high-throughput structural biology, active learning, and community challenges to build mechanistically grounded predictive models.
Read More →An open science effort to improve prediction of safety and toxicity for small molecules through high-quality data, mechanistic insight, and machine learning.
Benchmarking activity and structure prediction on a large dataset of human PXR-active compounds, with both an activity track and a structure track.
Predictive models and experimental datasets from OpenADMET blind challenges and data generation efforts.
Examining how cofolding methods (Boltz-2 and OpenFold3) handle protein-ligand structure prediction for ADMET targets — and where they fall short.
Science seminars, challenge webinars, and workshop recordings from the OpenADMET community.
Meet the OMSF staff, Governing Board, and our collaborators from Octant and UCSF.
A perspective on how blind challenges can help the field honestly evaluate and advance predictive modeling in drug discovery.
