Ran Boltz-2 100 million times to simulate cell biology
dom_beaini · x · 2026-08-26
Dominique Beaini shares results from running Boltz-2 100 million times to simulate cell biology. The experiment generated an "affinity-prints" dataset across 9,000 targets and 500k ligands. Combined with Recursion's phenomics data, he built a minimal Virtual Cell with just 6 parameters that estimates phenomics similarity, and introduced a Lig2Cell score to assess ligand efficacy.
Related event: Researcher Runs Boltz-2 100 Million Times to Simulate Cell Biology(2 posts)→
More from Research
- PrimeIntellect verifiers v0.3.1: Model Interception and Persistent ACP Sessions — xeophon · 2026-08-27
- RAG Isn't Dead: Navigating Retrieval vs. Agentic Search — hugobowne · 2026-08-27
- Podcast: What happens when you let an AI run a science lab — JMarty97 · 2026-08-27
- AI audit not infallible: Refine missed a known lemma error in paper — littmath · 2026-08-27
- Mathematician Litt: Paper Errors Mostly From Badly Propagated Edits, Not Deep Flaws — littmath · 2026-08-27
- Terence Tao ran all his published papers through AI error-finding scaffolds — littmath · 2026-08-27