Microsoft Research says Quine, its experimental biology AI system, helped a Broad Institute team narrow thousands of compounds to a shortlist for pancreatic cancer cell experiments. The search and candidate selection took one weekend, the company says.
The team studied pancreatic ductal adenocarcinoma, where cancer cells can occupy distinct transcriptional states: patterns of genes active inside a cell. Those states have been associated with differences in how tumors respond to treatment. Microsoft frames cell state as a possible therapeutic target, but reports no evidence that either shift improves treatment response. The researchers asked whether compounds could move cells from a "classical" state toward a "basal" one, and whether the change could be reversed.
Quine’s model links evidence about gene sequences, proteins, chemicals, cell states and images. Its software connects the model to research tools, published literature and scientists. The team used Quine to rank candidate compounds before lab tests; experimental results then help guide what researchers investigate next.
Across multiple lab tests, Microsoft reports, the highest-ranked compounds produced the largest intended classical-to-basal shifts. Some of the strongest effects came from compounds with mechanisms the researchers did not expect.
The reverse shift, from basal to classical, proved weaker, as Quine had predicted. Quine also predicted that some compounds would move cells toward a distinct third phenotype—a different cell profile beyond the classical-basal pair. Experiments detected it, which Microsoft says points to a richer pancreatic cancer cell-state landscape than a simple two-state axis.
Microsoft's announcement does not name the compounds or explain the ranking criteria. It also omits assay and sample counts and the measured size of the shifts, leaving readers without the data needed to judge the result's magnitude or repeatability. These are company-reported cell experiments, not evidence that any compound benefits patients. Quine remains experimental, with initial access limited to selected research collaborations and a small fellows cohort.
Sources
- Microsoft Research: Introducing Quine and its pancreatic cancer experiments
- Quine project overview and research fellowship
- Image: Ed Uthman's pancreatic adenocarcinoma cytology image, CC BY 2.0; illustrative and not from Quine experiments.
.jpg){kind=link}