MIT's IRIS model decodes cell-to-cell signaling history across cell types
MIT News AI · rss · 2026-10-16
A team led by MIT's Pulin Li (Whitehead Institute) published in Nature Methods that each signaling pathway leaves a unique "fingerprint" in gene activity, and these fingerprints are consistent across cell types—overturning the assumption that every pathway must be mapped separately in each cell type.
They built IRIS, a neural-network model that infers which signals a cell received at different developmental stages from its overall gene activity, generalizing to cell types it never saw (akin to transfer learning). Trained on thousands of human embryonic stem cells exposed to combinations of six major pathways, IRIS accurately predicted when and where pathways activate in mouse gastrulation cells fated to become heart, gut, muscle, and spinal cord tissue.
The team also used IRIS to predict signals driving a lung-development-critical cell type, confirmed in mouse embryos. The work could accelerate stem cell engineering and organoid creation for studying diseases like asthma, lung cancer, and pulmonary fibrosis.
More from Research
- DIY benchmark ranks 18 models on subjective judgments, GLM-5.3 tops consensus match — arkuto · 2026-10-16
- UCSD Researchers Propose Dataset Watermarking Detectable via Black-Box API Access — yuxiangw_cs · 2026-10-16
- EmbodiedEvo tops RoboDojo as agents inherit reusable skills and lessons — Aiden_Tech_Ai · 2026-10-16
- CHI reviewer laments: papers' Discussion sections now cite nothing — IanArawjo · 2026-10-16
- Casper: Understanding 90% of neurons ≠ understanding the network — StephenLCasper · 2026-10-16
- NVIDIA Launches cuPotential in CUDA-X to Bring AI Agents to Semiconductor Materials Simulation — tomaszbednarz · 2026-10-16