Microsoft's Quine picked compounds that changed cancer cells

Microsoft introduced Quine on 29 September 2026, an experimental AI system that helps biologists decide which ideas to test in a laboratory. Researchers from Microsoft and the Broad Institute used it to rank compounds that might alter the state of pancreatic cancer cells. Several highly ranked candidates produced the intended shifts in laboratory assays, according to Microsoft Research.

The result is about choosing promising experiments. It is an early research finding in cells grown for laboratory work, with further scientific validation needed before any clinical use.

Quine helps scientists decide what to test

Quine combines a model of biology with tools that connect research questions, published evidence and laboratory feedback. A scientist can ask a question, compare proposed experiments and feed measurements from the lab into the next round of work. The system is being developed to draw on information about genes, proteins, chemistry, cells and images, rather than treating each kind of data as an isolated problem, Microsoft says.

That design addresses a practical bottleneck. A lab can test only a fraction of the possible compounds or interventions. Ranking candidates in software gives researchers a more focused set of experiments to run, while the measurements show which predictions survive contact with biology.

The pancreatic cancer experiment tested cell state

Microsoft and Broad Institute researchers used Quine to prioritise thousands of compounds for pancreatic ductal adenocarcinoma, the most common form of pancreatic cancer. They looked at whether compounds could move tumour cells between states associated with different behaviour and treatment response. The Project Ex Vivo collaboration studies these cell states using computational models alongside experimental cancer systems.

For one transition, from a classical to a basal cell state, Quine’s highest-ranked compounds produced the largest intended shifts across the assays Microsoft reported. The researchers also observed cells moving towards a third state that Quine had predicted. Microsoft says the process of narrowing thousands of candidates to a handful for testing took a weekend.

The direction of a cell-state change is an experimental observation. The work has yet to establish a treatment for patients, and a compound’s effect in a lab assay cannot establish clinical benefit. Earlier published research on pancreatic cancer models also shows that cell states vary with the experimental system used, making follow-up tests especially important.

Researchers will get limited access through a fellowship

Microsoft opened applications for a 16-week Quine research fellowship on 29 September 2026. Its programme page says selected researchers will work with Microsoft scientists and receive access to the system and computing resources. The advertised research areas include protein and enzyme design, changes to cell state, and early-stage therapeutic research. Applications close on 2 November 2026, with the fellowship scheduled for June to September 2027.

Quine is at a research stage. The next useful evidence will come from other scientists testing whether its rankings improve their own experiments and whether the reported cell-state effects hold across more biological models.