Machine learning of genomic features in organotropic metastases stratifies progression risk of primary tumors

Biaobin Jiang,Quanhua Mu,Fufang Qiu,Yong Cao,Weiqi Xu,Weilun Fu,Jun Yu,Xuefeng Li,Jiguang Wang

doi:10.1038/s41467-021-27017-w

Abstract

Metastatic cancer is associated with poor patient prognosis but its spatiotemporal behavior remains unpredictable at early stage. Here we develop MetaNet, a computational framework that integrates clinical and sequencing data from 32,176 primary and metastatic cancer cases, to assess metastatic risks of primary tumors. MetaNet achieves high accuracy in distinguishing the metastasis from the primary in breast and prostate cancers. From the prediction, we identify Metastasis-Featuring Primary (MFP) tumors, a subset of primary tumors with genomic features enriched in metastasis and demonstrate their higher metastatic risk and shorter disease-free survival. In addition, we identify genomic alterations associated with organ-specific metastases and employ them to stratify patients into various risk groups with propensities toward different metastatic organs. This organotropic stratification method achieves better prognostic value than the standard histological grading system in prostate cancer, especially in the identification of Bone-MFP and Liver-MFP subtypes, with potential in informing organ-specific examinations in follow-ups.

Highlights

Metastatic cancer is associated with poor patient prognosis but its spatiotemporal behavior remains unpredictable at early stage
We showed that compared to the baseline models trained only by the clinical and histological features without the genomic data, the genomics-based model can accurately identify the metastatic tumors from the primary in the breast and prostate cancers (Area Under the Receiver Operating Characteristic curve, AUROC > 0.8), rather than in the lung and colon cancers (Fig. 2c and Supplementary Fig. 2b)
Further examining each ESR1 mutation in the breast cancer samples (n = 405, 129 from MSK, 262 from Foundation Medicine Inc. (FMI), and 14 from MET500), we found that the liver metastasis enrichment is primarily contributed by four hotspot positions located at the ligand-binding domain (LBD): D538, Y537, L538 and E380 (n > 20, Fig. 5c), all of which are spatially close to each other and have been demonstrated to give rise to estrogen-independent activation of downstream signaling and promote cellular proliferation[43]

Summary

Introduction

Metastatic cancer is associated with poor patient prognosis but its spatiotemporal behavior remains unpredictable at early stage. We identify genomic alterations associated with organ-specific metastases and employ them to stratify patients into various risk groups with propensities toward different metastatic organs This organotropic stratification method achieves better prognostic value than the standard histological grading system in prostate cancer, especially in the identification of Bone-MFP and Liver-MFP subtypes, with potential in informing organ-specific examinations in follow-ups. We aim to develop a Metastatic Network model (MetaNet) to assess metastatic risk and potential destination organs through collecting and analyzing a total of 32,176 pancancer DNA-sequencing samples Using this big-data cohort, we identify genomic biomarkers associated with common and organotropic metastases and validate their utility in metastatic risk assessment at an early stage using a machine-learning model to sort out a distinguishing subtype, namely Metastasis-Featuring Primary, with shorter disease-free survival than Conventional Primary patients. To facilitate the metastatic risk assessment and other organotropic biomarkers validation, we develop a web application of MetaNet which is available at https://wanglab.shinyapps.io/ metanet

Objectives

Methods

Results

Conclusion