Microsoft Introduces Quine, an Experimental AI System for Biological Research
Microsoft Research introduced Quine, an experimental artificial intelligence system designed to model biology and connect scientific tools, data, and researchers. The system was recently used with partner researchers to prioritize compounds that might shift tumor states in pancreatic cancer laboratory tests.
In short: Microsoft Research introduced Quine, an experimental artificial intelligence system designed to model biology and connect scientific tools, data, and researchers. The system was recently used with partner researchers to prioritize compounds that might shift tumor states in pancreatic cancer laboratory tests.
Biology is wonderfully complex, but traditional laboratory experiments take a long time and face limits in what they can explore at once. Now, computer scientists are trying to build artificial intelligence systems that can help researchers map out biological puzzles much faster than before.
What happened, in plain words
Microsoft Research announced Quine, a new experimental research effort created to work across the many different boundaries of biology. Developed in collaboration with researchers at the Broad Institute of Harvard and MIT, the system combines a biological world model with an interactive harness that links scientific tools, literature, wet labs, and human researchers. In practical tests involving pancreatic cancer cells, the system was used to predict and prioritize thousands of compounds based on their ability to shift tumor cells between different cellular states. Top-ranked compounds were then tested and validated in wet-lab assays, successfully narrowing down options over a single weekend.
Key points
- Connecting multiple scales of biology Quine learns shared representations across various biological scales and modalities, including gene sequences, protein structures, cellular states, and imaging data, so that evidence from one area can inform predictions in another.
- Testing real cancer biology in the lab Working with the Broad Institute of Harvard and MIT, researchers used Quine to identify compounds predicted to shift pancreatic cancer cells between different treatment-responsive states, validating the top candidates in wet-lab studies.
- Accelerating the scientific feedback loop The system aims to help scientists use computation to explore, propose, and rank potential paths forward before committing scarce laboratory resources, creating a continuous loop where experiments improve future model predictions.
- Controlled access through the Quine Fellows program Microsoft is launching the Quine Fellows program to give a select cohort of scientists access to the experimental technology to accelerate their own research and provide feedback, with plans to eventually expand access through products like Microsoft Discovery.
Terms explained
- Multimodal world model — A computer system that learns from many different types of information at the same time to simulate how a complex system works. Example: Learning about an animal by combining its genetic sequence, its physical bone structure, and pictures of its cells all in one place.
- Wet-lab assays — Physical experiments conducted in a laboratory using liquids, chemicals, and biological materials rather than purely on a computer. Example: Testing a liquid medicine directly on living cancer cells grown in a Petri dish.
- Cellular state — The specific condition or behavior of a cell at a given time, which can influence how it reacts to treatments. Example: A cell shifting from a resting phase to an active phase where it fights off a medicine differently.
Why it matters
By helping scientists narrow down huge lists of potential drug compounds in a fraction of the time, tools like Quine could eventually help researchers save months of experimental work and reduce research costs when looking for new medical treatments.
What we still don't know
Quine is strictly an experimental research technology meant for research use only, not clinical or medical treatments. Its outputs can be incomplete or inaccurate, and they require careful review and validation by qualified researchers. The current technology is still maturing, and the reverse state transition for cancer cells proved much more difficult for the model to predict.
Based on reporting from Microsoft Research. This is an independent explainer, written in our own words with AI assistance; Microsoft Research has not reviewed or endorsed it. Read the original for the full details.