Data in Biotech Success-Driven Drug Discovery with OpenBench CEO James Yoder
12 snips
Feb 11, 2026 James Yoder, Founder and CEO of OpenBench and former statistician turned ML leader, describes a success-driven approach that charges only for validated hits. He discusses ultra-large virtual screening, computational pipelines for binding prediction, and how data flywheels and active learning improve hit discovery. He also explains risk-sharing collaborations and real-world case studies.
AI Snips
Chapters
Transcript
Episode notes
Price Projects By Quantifying Epistemic Risk
- Assess epistemic scientific risk and translate it into a single quoted price before committing to a success-driven deal.
- Adjust pricing and risk appetite case-by-case to keep expected value profitable while offering clients fair terms.
High Reported Success Rate
- OpenBench reports a 75% success rate across 16 completed commercial projects, often delivering multiple series per success.
- They target ~80% a priori success probability but will price and accept lower-probability projects if economics permit.
Active-Learning Virtual Screening Pipeline
- Their computational pipeline samples ultra-large virtual libraries, docks sampled ligands into chosen protein snapshots, then trains proxy models to search reaction subspaces.
- They run an active-learning loop: dock+score to build training labels, train proxy, infer across space, rescore, and iterate until saturation.