Learning chemical sensitivity reveals mechanisms of cellular response

This paper introduces ChemProbe, a deep learning model that predicts cellular sensitivity to molecular probes and drugs by integrating transcriptomic and chemical structural data. ChemProbe accurately predicts drug response in cancer cell lines, tumor samples, and retrospectively evaluates precision breast cancer treatment. Prospective validation in new cellular models confirms its accuracy. Model interpretation reveals transcriptomic features reflecting compound targets and protein network modules, identifying genes driving ferroptosis. ChemProbe is presented as an interpretable *in silico* screening tool for research into molecular mechanisms of chemical sensitivity.

William Connell, Kristle Garcia, Hani Goodarzi, Michael J. Keiser