Tahoe-100M: Mapping drug-induced molecular phenotypes at single-cell resolution

Tahoe-100M is a publicly available atlas of 100 million single-cell transcriptomes that captures the molecular responses of 50 cancer cell lines to approximately 1,100 drug-dose combinations, creating an unprecedented resource for linking drug mechanisms to cellular phenotypes. By pairing single-cell resolution with large-scale pharmacological perturbation, the dataset enables systematic discovery of how drugs reshape gene-expression programs across diverse cancer backgrounds. The openly released resource is explicitly designed as a training substrate for predictive AI models of cell behavior, bridging high-throughput screening with machine-learning-driven drug discovery.