Microsoft Research introduced Quine on Tuesday, a system that combines a world model of biology with a harness connecting models, scientific tools, literature and the researchers using them.
The premise is that biology resists the boundaries models usually draw. Genes influence proteins, proteins act inside cells, cells form tissues, and each experiment reshapes the next question. Quine learns shared representations across sequence, structure, function, cellular state and imaging data, so evidence from one modality can inform predictions in another.
Microsoft describes the world model as something that represents the state of a biological system, predicts how that state changes under intervention, and reasons several steps ahead. The goal is not to replace the wet lab but to narrow what deserves scarce bench time.
The company pointed to work on pancreatic ductal adenocarcinoma with researchers at the Broad Institute of MIT and Harvard. Quine was asked to rank thousands of compounds by their likelihood of shifting tumor cells between therapeutically relevant states. Its highest-ranked candidates produced the largest intended shifts in wet-lab assays, and some of the strongest effects came from compounds with unexpected mechanisms. The whole narrowing process took a weekend.
Microsoft is also opening a Quine Fellows program to give outside scientists access. It labels the system experimental research technology, not for clinical use, and says outputs need review by qualified researchers.
