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

Area of Science:

• Oncology
• Genomics
• Computational Biology

Background:

• Understanding the cell of origin (COO) for various cancers is crucial for developing effective diagnostic and therapeutic strategies.
• Somatic mutations in cancer's COO tend to accumulate in closed chromatin regions, providing a potential marker for identification.

Purpose of the Study:

• To predict the cell of origin (COO) for 37 cancer subtypes using a data-driven approach.
• To confirm known cancer origins and identify specific cell subsets, enhancing our understanding of tumor development.

Main Methods:

• Integration of 3,669 whole genome sequencing patient samples with 559 single-cell chromatin accessibility profiles.
• Application of machine learning algorithms to predict the COO based on genomic and epigenetic data.

Main Results:

• High accuracy and robustness in predicting the COO for 37 cancer subtypes, validating known origins.
• Identification of a basal COO for small cell lung cancers and a neuroendocrine COO for atypical cases.
• Uncovered distinct cellular trajectories and an intermediate metaplastic state in tumorigenesis for gastrointestinal cancers.

Conclusions:

• The study provides a robust computational framework for identifying cancer COO.
• Findings offer critical implications for cancer prevention, early detection, and personalized treatment stratification.
• Highlights the utility of integrating multi-omics data for deciphering complex cancer biology.