IRIS reconstructs cell-signaling histories via perturbation screens and transfer learning
Dorian182 · reddit · 2026-10-19
IRIS, published in Nature Methods, infers cell-signaling states and histories across cell types from perturbation screens and in vivo scRNA-seq data.
- Core idea: combine combinatorial ligand perturbation screens (in vitro) with in vivo single-cell measurements, using transfer learning to map learned signaling responses onto in vivo cells
- Architecture: a conditional VAE encodes a latent signaling space, with a classifier neural network handling label transfer; the two are trained iteratively
- Validation: in-sample tests on mES cell screens for TGF-β, WNT, FGF, BMP, and RA pathways show ROC/PR curves clearly beating a response-gene baseline
- Applications: infers binary signaling states (e.g., Notch, MAPK) in animal, organ, and stem cells, supporting developmental biology and drug research
A notable AI-for-Science advance in single-cell signaling inference.
More from Research
- Top probability journals shut submissions for a month citing surge and AI impact — michaelchchoi · 2026-10-19
- Bare IPython Harness Matches Production Agents on Qwen3.5-9B; Output Normalization Kills Run-to-Run Variance — HauntingMoment · 2026-10-19
- Bryan Kelly and Dacheng Xiu release 159-page Financial Machine Learning textbook free — mdancho84 · 2026-10-19
- Six lessons from 60 papers: what a week of AI-for-Science research reveals — bravo_abad · 2026-10-19
- From a 2014 MSR bet to Nature: how deep learning conquered weather forecasting — LesterMackey · 2026-10-19
- Amateurs Are Using GPT-6 to Crack Decades-Old Ciphers, Overwhelming Cryptographers — ChuckDBrooks · 2026-10-19