Researchers at MIT and the Whitehead Institute have developed an artificial intelligence model that can read the chemical messages cells use to talk to each other. The tool is called IRIS.
The study was led by Pulin Li, an MIT assistant professor of biology and Whitehead Institute member, and graduate student Nicholas Hutchins. Their findings were published in the journal Nature Methods on Sept. 8.
As an embryo grows, stem cells turn into specialized cells such as brain, liver, or muscle cells. What each cell becomes depends on which of its genes are switched on or off.
Cells do not make this choice alone. They send and receive chemical signals from nearby cells through multi-step sequences called signaling pathways.
How the IRIS Model Works
Scientists had long believed that signaling pathways worked differently in each cell type. That meant they would need to map every pathway in every cell type, one at a time.
The new study found that each pathway leaves behind a unique “fingerprint.” This is a pattern of gene activity that shows which signals a cell has received.
These fingerprints stay the same across different cell types for the same pathway. This lets researchers trace signaling histories across many cell types at once.
IRIS is a neural network, a type of AI system built to find patterns in complex data. It looks at a cell’s overall gene activity and estimates which pathways were active at certain times.
Li compared it to voice assistants like Siri, which are trained mostly in English but can learn other languages. She said this method, called transfer learning, is why IRIS works across many cell types.
The team trained IRIS on data showing how thousands of human embryonic stem cells reacted to dozens of combinations of six major signaling pathways. This created an atlas of how signals shape cell behavior.
Testing on Mouse Embryos and Lung Cells
The researchers then tested IRIS on single cells from mouse embryos during gastrulation. This is a stage when cells quickly branch into different fates.
IRIS correctly predicted when and where certain pathways switched on in cells headed to become heart, gut, muscle, and spinal cord tissue.
The team also used IRIS to find signals needed to create a cell type important for lung development. The model predicted that turning on a specific pathway would push cells toward lung development, and experiments in mouse embryos confirmed it.
“IRIS is helping us decode the language cells use to talk to each other at a much faster rate than we could realistically achieve through experiments,” Hutchins said.
The researchers said the findings could speed up stem cell engineering and improve organoids, which are small 3D models that mimic real organs.
Better lab-grown lung models could help scientists study asthma, lung cancer, and pulmonary fibrosis, a disease that scars lung tissue. Researchers said these models could be used to test therapies and design treatments to repair damaged tissue.
