769-Benchmarking Tumor Subclonal Reconstruction Algorithms

This article details the results of a seven-year crowd-sourced benchmarking challenge designed to evaluate algorithms used for reconstructing tumor evolution. Researchers tested 31 different computational methods against 51 simulated cancer genomes to see how accurately they could identify subclonal populations and ancestral relationships. Interestingly, no single method outperformed all others across every category, and ensemble strategies failed to improve upon the best individual tools. The authors have provided their containerized methods and datasets as an open resource to help the scientific community develop more refined tools for understanding how cancers progress.

References:

• Salcedo A, Tarabichi M, Buchanan A, et al. Crowd-sourced benchmarking of single-sample tumor subclonal reconstruction[J]. Nature biotechnology, 2025, 43(4): 581-592.