Direct discovery · Biology · Medicine
Quine prioritizes compounds that produce predicted pancreatic tumor-state shifts
Wet-lab-validated compound rankings for shifting pancreatic cancer transcriptional states, including an experimentally observed third phenotype predicted by the system.
Summary
Microsoft Research and Broad Institute collaborators used Quine to prioritize compounds predicted to shift pancreatic cancer cells between therapeutically relevant transcriptional states. Top-ranked compounds produced the largest intended shifts across wet-lab assays, and several predicted a distinct third phenotype that was also observed experimentally. Microsoft reports that narrowing thousands of compounds to a testable shortlist took one weekend.
AI role
A multimodal biological world model and research harness ranked thousands of compounds for their predicted effects on pancreatic ductal adenocarcinoma cell states before laboratory testing.
Narrative role
Provides a concrete closed-loop biology example in which a multimodal system generated experimentally testable rankings and an unexpected phenotype hypothesis, rather than only launching a platform or reporting benchmark scores.
Caveat
The evidence is a first-party research announcement without a linked paper, candidate identities, assay counts or a controlled time comparison. The reported months saved are prospective rather than independently measured.