Active learning scans 2.5M molecules to find new way to block RAS-driven cancers
bravo_abad · x · 2026-08-28
RAS mutations drive roughly a quarter of human cancers, yet many clinically important RAS variants remain hard to target directly. Yoko Yoshikawa and coauthors used active learning to search for a different point of attack.
Key design points:
- Starting from a commercial library of 2.5 million molecules, with a training set of just 7 known active compounds and 1,493 inactive ones;
- Instead of docking every molecule to a fixed protein structure, their ligand-based active-learning system iteratively selects compounds for experimental testing and feeds results back into the model;
- Crucially, molecules structurally similar to previous RAS inhibitors are deliberately excluded, pushing the search into novel chemical space.
More from Research
- PySR v2 Test: Recovers CAD Models from Point Clouds, Discovers Shaders from Data — MilesCranmer · 2026-08-28
- TerminalBench-Science v0.1: Opus 5 leads overall benchmarks — BenBlaiszik · 2026-08-28
- TerminalBench-Science sets high bar, slashing model scores by 10+ points — BenBlaiszik · 2026-08-28
- OpenResearch Launches AutoResearch: Automating Paper Replication with Agent Swarms — simonguozirui · 2026-08-28
- Auto-research loops are the future: RL discovers novel states 5x more efficiently — const_reborn · 2026-08-28
- Intrinsic Discovery method enables unsupervised model exploration — burny_tech · 2026-08-28