MIT's IRIS is learning the language cells speak to decide what they will become

A cell deciding whether to become heart, gut or lung is listening to chemical conversations — signals that switch genes on and off in patterns. MIT and Whitehead Institute researchers have trained a machine-learning model, IRIS, on thousands of human embryonic stem cells exposed to dozens of signal combinations, and found that each signal leaves a unique gene-activity 'fingerprint' that stays consistent across cell types. Published in Nature Methods, IRIS then predicted signalling in mouse embryo cells bound for heart, gut, muscle and spinal cord — and its lung prediction held up in real mouse experiments. 'IRIS is helping us decode the language cells use to talk to each other,' says graduate student Nicholas Hutchins. The hope: better stem-cell engineering and lab-grown tissue.

SDG 9 · Industry, innovation and infrastructure — Mechanism: tech innovation