Microsoft Research unveils Quine, a multimodal 'world model of biology' validated in pancreatic-cancer screens with the Broad Institute
On Sept 29, 2026 Microsoft Research introduced Quine, an early-stage AI research system that learns shared representations across genomics, proteins, chemistry, cell state and bioimaging and is wired into a tool/literature harness. With the Broad Institute it prioritized thousands of compounds to shift pancreatic-cancer (PDAC) cell states in about a weekend, and Microsoft says the top-ranked candidates produced the largest intended transcriptional shifts in the lab. Access is limited to a Quine Fellows program.
Key facts
- Announced Sept 29, 2026 by Nicolo Fusi (VP, Distinguished Scientist) and Jonathan M. Carlson (VP), Microsoft Research
- Multimodal model across genomics, proteins, chemistry, cellular state and bioimaging, plus a harness connecting tools, literature and researchers in a closed loop with wet-lab experiments
- PDAC case study with the Broad Institute: thousands of compounds predicted and prioritized for cell-state transitions in about one weekend; top candidates gave the largest intended transcriptional shifts; predictions pointed to a third phenotype beyond the classical/basal axis and to repurposing candidates (company claims, no paper linked)
- Access: Quine Fellows program (applications open) and selected collaborations; planned to reach customers through Microsoft Discovery; research use only
Science result
- Field
- biology / cancer biology / drug discovery
- Problem
- Prioritizing compounds that shift pancreatic ductal adenocarcinoma cell states
- Result
- AI-ranked compounds produced the largest intended transcriptional shifts in validation experiments (per Microsoft)
- AI system
- Quine
- Human role
- Human-led with AI tools: scientists pose questions, Quine proposes, Broad Institute labs validate
- Verification
- lab-validated (company-reported; no peer-reviewed paper yet)
- Status
- pending
What happened
Microsoft Research presented Quine as a step toward a "world model of biology" that lets scientists search far more interventions computationally than intuition allows, then test the best in the lab. The only result shown is a PDAC compound-prioritization study with the Broad Institute; no paper, benchmark or weights were released.
Why it matters
It joins a crowded field of lab-in-the-loop biology agents (FutureHouse/Edison Robin, Lila, OpenAI GPT-Rosalind) and is Microsoft's first named biology foundation-model effort tied to its Discovery platform.
Changelog
- 2026-10-05: created (found via Microsoft Research RSS during the 06:30 run)
Related events
Sources (1)
id: 2026-09-29-microsoft-quine-biology-world-model · updated 2026-10-05 · open in the interactive timeline