Reconstructing phylogenies of metastatic cancers

Abstract

Reconstructing the evolutionary history of metastases is critical for understanding their basic biological principles and has profound clinical implications1–3. Genome-wide sequencing data has enabled modern phylogenomic methods to accurately dissect subclones and their phylogenies from noisy and impure bulk tumor samples at unprecedented depth4–7. However, existing methods are not designed to infer metastatic seeding patterns. We have developed a tool, called Treeomics, that utilizes Bayesian inference and Integer Linear Programming to reconstruct the phylogeny of metastases. Treeomics allowed us to infer comprehensive seeding patterns for pancreatic8, ovarian9, and prostate cancers10,11. Moreover, Treeomics correctly disambiguated true seeding patterns from sequencing artifacts; 7% of variants were misclassified by conventional statistical methods. These artifacts can skew phylogenies by creating illusory tumor heterogeneity among distinct samples. Last, we performed in silico benchmarking on simulated tumor phylogenies across a wide range of sample purities (30-90%) and sequencing depths (50-800x) to demonstrate the high accuracy of Treeomics compared to existing methods.