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

Hutchins et al. (Nature Methods 2026): IRIS, a neural network trained on a signaling-perturbation atlas of human pluripotent stem cells, infers signaling activities and histories of in vivo cell types; applied to mouse-embryo single-cell atlases. Developmental biology, not a tumor study.

Authors / organisation

Hutchins NT, Meziane M, Lu C, Mitalipova M, Fischer DS, Li P.

Venue

Nature Methods

Date

2026-09-08

Type

paper

Year

2026

DOI

10.1038/s41592-026-03213-8

PMID

42711495

Verification

Crossref (title, 6 authors, 2026-09-08, Nature Methods 23(10):1973-1983) + Europe PMC (PMID 42711495, abstract); MIT News 2026-10-15 corroborates Sept 8 publication; full text not read

signalingtransfer-learningperturbation-atlasstem-cellsdevelopment

Open in the searchable library → · JSON