MintCNA: A Unified Framework for Integrative Copy Number Profiling with Single-Cell Multi-Omics Data

Abstract

Chromosomal copy number alterations (CNAs) are key drivers of tumor evolution, disease progression and therapeutic resistance, and their identification is critical for delineating tumor clonal structure. However, accurately detecting CNAs from single-cell data remains challenging as most existing CNA callers still analyze a single omics layer and are susceptible to assay-specific noise. Recently, single-cell multi-omics techniques have enabled joint sequencing of multiple molecular layers (e.g., genomics and transcriptomics) simultaneously in the same cells, yet computational methods that fully exploit such complementary multi-omics data for CNA analysis remain limited. Here, we present MintCNA, a unified framework for CNA detection from paired single-cell multi-omics data. MintCNA integrates deep-learning based denoising with a missingness-adjusted multivariate change-point detection procedure to enhance CNA profiling across omics layers. In simulations, MintCNA outperformed existing single-cell CNA callers across all evaluated scenarios. MintCNA showed the highest correlation with bulk whole-genome sequencing in a colorectal cancer dataset and revealed progressive loss of chromosome 17 during metastasis. In an IDH-mutant glioma cohort, MintCNA generalized to paired single-cell ATAC-seq and RNA-seq data and identified copy number subclones corresponding to distinct malignant cell states. MintCNA provides an advanced multi-omics integrative single-cell CNA analysis tool for better delineating the mechanisms underlying disease progression and tumor evolution.