Designing the Strategy to Defeat Cancer

Zhang et al., Preprint. 2025

The Tahoe-100M dataset, profiling over 100 million cells across 50 cancer cell lines and 1,100 perturbations to build predictive models of cell behavior. By growing cell lines in spheroids and treating them with small molecules at varying concentrations, they minimized batch effects while creating the largest single-cell atlas to date. Using this dataset, they mapped drug responses by mechanism of action, revealing diverse cellular reactions even within drug classes. Further analysis of the RAS/RAF pathway demonstrated how genotype influences differential gene expression, offering insights into biomarkers and treatment resistance. This dataset paves the way for AI-driven biological predictions, enhancing drug discovery by not only determining if a drug works but also uncovering the mechanisms behind its effects.