AI model decodes cell signaling fingerprints, opens path to lab-grown organs

Whitehead Institute researchers have developed IRIS, a machine learning model that detects unique 'fingerprints' left by signaling pathways in cells. The model works across cell types, potentially accelerating stem cell engineering and organoid creation for drug testing and regenerative medicine.

Bottom line — The Nature Methods study shows IRIS can predict signaling histories across cell types, confirmed by lung development experiments in mouse embryos.

Go deeper (5)

• IRIS is a neural network trained on thousands of human embryonic stem cells responding to dozens of combinations of six major signaling pathways, according to the study published in Nature Methods.
• The team tested IRIS on single cells from mouse embryos during gastrulation, accurately predicting signaling activation in cells destined for heart, gut, muscle, and spinal cord tissue, the researchers said.
• When IRIS predicted that activating a specific pathway would drive lung development, experiments in mouse embryos confirmed the model's prediction, the study reports.
• The approach uses transfer learning — like a voice recognition system trained in English recognizing other languages, explained Pulin Li, assistant professor at MIT and Whitehead Institute member.
• Identifying the precise signals that guide stem cells could enable more reliable lab-grown lung tissue for studying asthma, lung cancer, and pulmonary fibrosis, the researchers noted.

Read the reporting