Inferring cancer type-specific patterns of metastatic spread using Metient

Abstract

Cancers differ in how they establish metastases. These differences can be studied by reconstructing the metastatic spread of a cancer from sequencing data of multiple tumors. Current methods to do so are limited by computational scalability and rely on technical assumptions that do not reflect current clinical knowledge. Metient overcomes these limitations using gradient-based, multi-objective optimization to generate multiple hypotheses of metastatic spread and rescores these hypotheses using independent data on genetic distance and organotropism. Unlike current methods, Metient can be used with both clinical sequencing data and barcode-based lineage tracing in preclinical models, enhancing its translatability across systems. In a reanalysis of metastasis in 169 patients and 490 tumors, Metient automatically identifies cancer type-specific trends of metastatic dissemination in melanoma, high-risk neuroblastoma, and non-small cell lung cancer. Its reconstructions often align with expert analyses but frequently reveal more plausible migration histories, including those with more metastasis-to-metastasis seeding and higher polyclonal seeding, offering new avenues for targeting metastatic cells. Metients findings challenge existing assumptions about metastatic spread, enhance our understanding of cancer type-specific metastasis, and offer insights that inform future clinical treatment strategies of metastasis.