A Pan-Cancer Single-Cell Compendium of Intratumoural Heterogeneity.

Abstract

Intratumoural heterogeneity remains one of the most formidable challenges in oncology, driving treatment resistance, disease progression, and poor patient outcomes. To uncover the cellular programs that define tumour composition and diversity, we interrogated over 3.6 million single cells derived from more than 1,000 primary and metastatic tumours spanning 14 diverse cancer types. Through the identification and quantification of coordinated gene expression programs, we delineated heterogeneous cancer cell states, exhibiting unique and clinically significant associations with patient survival. Integration with high-resolution spatial transcriptomics identified prognostic cancer cell states marked by aggressive transcriptional programs, co-localizing with invasive histological features and significantly enriched in metastatic settings. Functional drug perturbation screening uncovered drug vulnerabilities specific to distinct cancer cell states, supporting both direct targeting strategies and therapeutic reprogramming toward less aggressive cellular phenotypes. Collectively, this study offers novel insights into the relationship between intratumoural transcriptional heterogeneity, spatial organization, clinical outcomes, and therapeutic vulnerabilities.