Deep learning predicts patient drug responses from single cell data
scDEFT: A deep learning framework for drug-effect prediction and counterfactual reasoning
Machine Learning
Summary
Treatments can affect patients differently, and understanding why is important for better medicine. The authors introduce scDEFT, a deep learning tool that learns how drugs change individual cells, helping predict how patients will respond before treatment. It uses large datasets of millions of cells and recognizes patterns that standard methods miss. This helps suggest which treatments or targets might work best for different patients and can even imagine effects from new drug combinations.
What this means in practice
- •For clinical data scientists: Predict patient response to specific drugs before treatment using single cell data and improve patient stratification.
- •For pharmaceutical research teams: Identify gene drivers and potential co-targets by interpreting drug effects at single cell resolution across patient cohorts.
Authors
Murthy Devarakonda
Abstract
Longitudinal single cell atlases now capture matched pre treatment and post treatment states from responders and non responders, presenting an opportunity to mechanistically explain why two patients on the same drug diverge. We introduce scDEFT (single cell Drug EFfect Transducer), which treats a drug as a conditioning operator on cell representations, enabling prediction and explanation. In scDEFT, feature wise linear modulation produces drug conditioned cell latents, learned under abundant per cell supervision and then frozen. Two independent heads aggregate those latents over shared transcriptional neighborhoods to predict drug induced state change and responder status. A backward stage ranks the latent dimensions by how strongly they separate responders from non responders and maps them to genes under a cell composition control. On a harmonized inflammatory bowel disease atlas of 1.16 million cells, three cohorts and two drug classes, scDEFT predicts state change at 45% of the baseline to reproducibility ceiling headroom and stratifies responders before treatment at AUROC 0.70, where standard predictors remain at chance. These predictions and the drivers behind them support target and co target nomination, patient stratification, and counterfactual prediction of unseen drug cohort effects.