Acellera CEO runs an unsupervised AI-scientist lab for real molecular hit identification
gdefabritiis · x · 2026-10-21
Acellera CEO Gianni De Fabritiis published experiment #2 in his AI-lab series: a team of autonomous AI agents (built on PlayMolecule AI) drives a real drug discovery program with minimal human direction.
- He supplied only a target product profile (TPP) and resource limits for IL23–receptor interference (a psoriasis-relevant pathway); agents chose their own next steps.
- Progress was tracked via a shared wiki and a molecule leaderboard; agents explored structural evidence (PDB 6WDQ, 8UUI), proposed molecules, ran computational tests, and decided what to investigate next.
- Experiment #1 had shown better base models yield better TPPs and strategies; this one tests whether an AI lab can do hit identification end-to-end.
- Full project report PDF, blog, white paper, and a GitHub repo of AI-lab notes are linked.
Related event: Acellera Shows Autonomous AI Agent Lab Can Find Drug Leads(2 posts)→
More from coding & agent
- Beat AI's samey frontends: import a design system before generating UI — dotey · 2026-10-21
- Trust slider over Clairic: 102 manual reviews catch 74 of 76 extraction errors — michellechen · 2026-10-21
- FreePi Bets on Ad-Supported Coding Help to Undercut $20/Month AI Coding Subscriptions — heyneighbor · 2026-10-21
- RedMonk: MCP and A2A Verify Publishers, Not Safety — Registries Become the Trust Layer — rseroter · 2026-10-21
- Agentic refactor: 600 PRs moved OpenClaw's web UI from Lit to Solid 2.0 in 2 weeks — steipete · 2026-10-21
- 19 Recorded Runs of 18 Agents Building a Lisp Interpreter, Released on Hugging Face — strickvl · 2026-10-21