Consensus Copy-Number Alteration Signatures from Clinical Panels Enable Pan-Cancer Risk Stratification and Therapy Response Association

TL;DR

This study introduces a method to extract meaningful CNA signatures from routine cancer panel data, enabling better cancer risk assessment and treatment predictions.

Contribution

A novel consensus framework integrating four deconvolution algorithms to extract CNA signatures from sparse clinical panel data.

Findings

Five reproducible CNA signatures (CON1–CON5) were identified across 24,870 tumors.

The signatures correlated with overall survival and therapy resistance across multiple cancer types.

Associations with driver mutations like GATA3 and KRAS confirmed biological relevance.

Abstract

Somatic copy-number alterations (CNAs) are pervasive in cancer, but routine targeted panels yield sparse CNA readouts unsuited for CNA signature analysis. We built a consensus framework that integrates four deconvolution algorithms to extract CNA signatures from panel data. Analysis of 24,870 tumors sequenced using MSK-IMPACT identified five reproducible signatures (CON1–CON5). CON5 mirrored near-diploid profiles, whereas the others captured distinct aneuploid patterns. Technical fidelity was confirmed by internal cross-validation and external validation in sarcoma and hepatocellular carcinoma cohorts. Clinically, these signatures were associated with overall survival across tumor types (hazard ratio 1.3–2.5; FDR < 0.01) and provided additive prognostic information beyond Fraction of Genome Altered. Associations with driver mutations (GATA3 in CON1, KRAS in CON5) supported biological specificity, and the signatures delineated resistance landscapes for chemotherapy, hormonal, targeted, and immunotherapy. By converting routine panel data into biologically interpretable prognostic features, our framework enables risk stratification and therapeutic guidance in precision oncology.

Figures