Comprehensive Summary
This analysis by Bairakdar and colleagues examines the challenge of identifying the cell of origin (COO) across cancers, which is key for understanding tumor initiation and progression. The authors introduce a data-driven machine learning framework called SCOOP (Single-cell Cell Of Origin Predictor) that integrates 3,669 whole-genome sequencing samples with 559 single-cell chromatin accessibility profiles to predict the COO across 37 cancer subtypes. The framework achieves high robustness; for example, it pointed towards basal cells as an origin point in small-cell lung cancers, highlighted metaplastic intermediate states in gastrointestinal cancers, and also showed that fetal-like multipotent progenitor cells are a source of gliomas. These findings show cancer-specific cellular trajectories and provide detailed COO maps across multiple tumor types.
Outcomes and Implications
The work of SCOOP advances the precision of cancer classification, helping to improve methods of prevention, early detection, and treatment. By mapping COO at higher resolution, this framework can refine diagnostic biomarkers and support targeted therapy decisions. For certain cancers, identifying distinct cellular origins may guide treatment selection and surveillance strategies, while helping to find metaplastic states could point toward intervention windows. Overall, this approach strengthens the translational link between single-cell biology and clinical oncology, offering a foundation for more personalized and effective cancer management in the future.