Learning the cellular origins across cancers using single-cell chromatin landscapes.

Deciphering the pre-malignant cell of origin (COO) of different cancers is critical for understanding tumor development and improving diagnostic and therapeutic strategies in oncology. Prior work demonstrates that somatic mutations preferentially accumulate in closed chromatin regions of a cancer's COO. Leveraging this information, we combine 3,669 whole genome sequencing patient samples, 559 single-cell chromatin accessibility cellular profiles, and machine learning to predict the COO of 37 cancer subtypes with high robustness and accuracy, confirming both the known anatomical and cellular origins of numerous cancers, often at cell subset resolution. Importantly, our data-driven approach predicts a basal COO for most small cell lung cancers and a neuroendocrine COO for rare atypical cases. Our study also highlights distinct cellular trajectories during cancer development of different histological subtypes and uncovers an intermediate metaplastic state during tumorigenesis for multiple gastrointestinal cancers, which have important implications for cancer prevention, early detection, and treatment stratification.