Reconstructing signaling histories of single cells via perturbation screens and transfer learning.

Transferable signaling response signatures enable high-resolution reconstruction of in vivo cell signaling histories across diverse cell types.

• Why it matters: Understanding cellular signaling histories in vivo is crucial for manipulating cell states and guiding interventions, yet current high-throughput methods are lacking.
• What they did: The authors developed an integrated experimental-computational approach, training a neural network model called IRIS on a human pluripotent stem cell perturbation atlas with 1,000 experiments.
• The result: Applying IRIS to mouse embryo data revealed conserved signaling patterns, heterogeneity, and developmental trajectories, paving the way for targeted cell fate engineering and in vivo signaling mapping.