Phys.org — Biology · September 8, 2026

AI model decodes cell signaling fingerprints across diverse cell types

BiologyBiomedical Engineering

THE AI ANGLE

Predicting developmental signaling pathway activation across cell types using transfer learning

Researchers developed IRIS, a neural network model that identifies conserved gene activity 'fingerprints' left by signaling pathways across diverse cell types, overturning the assumption that pathways must be laboriously mapped for each cell lineage separately. By applying transfer learning to gene expression data, the model accurately reconstructed signaling histories during mouse embryonic gastrulation and identified the signals required to direct lung cell differentiation. This capability provides a practical roadmap for biomedical engineers and biologists to control stem cell fates, optimize organoid creation, and investigate disease mechanisms.

THE TEACHING ANGLE

Instructors can explore the shift from the traditional assumption that signaling pathways act in entirely cell-type-specific manners to a model where conserved transcriptional fingerprints enable cross-lineage transfer learning.

Read the original at phys.org Generate teaching or study materials

Instructors get discussion guides, assignments, and mini-cases. Students and readers get a plain summary, class prep, and an exercise. All built from the full article. Three are free with an account.

AI model decodes cell signaling fingerprints across diverse cell types

More in Biology