A knowledge-driven framework for predicting single-cell responses for unprofiled drugs

TL;DR - Feng et al. introduce MAP, a knowledge-driven AI framework for predicting single-cell responses to previously unprofiled drugs. Integrating biological mechanism knowledge improves generalization to untested compounds and supports virtual screening for cancer drug candidates.

• Predicts cellular responses to chemical perturbations at single-cell resolution.
• Incorporates biological mechanism knowledge rather than relying solely on observed perturbation data.
• Targets out-of-distribution generalization to drugs not profiled during training.
• Demonstrates potential for prioritizing cancer drug candidates through virtual screening.