Somatic copy number alterations (CNAs) are pervasive in cancer, particularly in tumours driven by chromosomal instability. Their strong association with disease progression and adverse clinical outcomes makes understanding how CNAs accumulate and change over time essential for early detection, prognosis, and personalized therapy.
Although numerous studies have characterized CNA patterns and relative timing, few have estimated CNA rates in absolute chronological time. We previously developed CNETML, a maximum-likelihood framework for reconstructing CNA-based phylogenetic trees and estimating CNA evolutionary rates from shallow whole-genome sequencing data. Applications of CNETML to cancer datasets have enabled quantitative reconstruction of tumour evolutionary histories. However, existing models typically assume a single class of CNA events and constant evolutionary dynamics, whereas cancer genomes often undergo multiple types of large-scale alterations that shape copy number evolution.
Here we extend CNETML with a multi-scale model that explicitly incorporates three classes of genomic events: segment-level duplications and deletions, chromosome-level gains and losses, and whole-genome doubling. These processes are represented by three transition matrices describing independent Markov chains for each alteration scale. The likelihood is computed dynamically by combining these matrices during phylogenetic inference, allowing different CNA processes to contribute jointly to genome evolution. To improve computational efficiency, the model prunes redundant haplotype copy-number states that are incompatible with the constraining maximum copy number boundaries, substantially reducing the size of the state space. Implemented within the CNETML framework, the new model enables more efficient and biologically realistic inference of tumour phylogenies and CNA evolutionary rates from cancer sequencing data.

Co-authors: Bingxin Lu