RECAM: Reconstructing Evolutionary dynamics of Copy number Alterations from Multiple samples across cancer types

Somatic copy number alterations (CNAs), gains or losses of large genomic regions, are widely prevalent in cancers especially those driven by chromosomal instability. CNAs are associated with disease progression and poor patient outcomes, such as copy number gain of oncogene KRAS in pancreatic cancer. Understanding how CNAs evolve over time during the progression from precancerous lesions to malignancy, metastasis, and treatment resistance is important to detect cancer earlier, predict cancer progression, and inform personalized treatment. Despite numerous studies on the patterns and timing of CNAs, few studies have managed to estimate CNA rates in chronological time, which will be addressed by this project. The declining DNA sequencing cost has increased accessibility to multiple samples from individual patients across space and/or time, which allows us to reconstruct evolutionary dynamics of CNA-driven cancer genomes. The latest development and clinical application of liquid biopsy enable minimally invasive genome-wide profiling of CNAs over time in a cost-effective way. Rapid ultra-low coverage copy number profiling of cell-free DNA has been proposed for precision oncology screening. The rise of longitudinal sequencing on patient-derived cell culture is also providing opportunities to monitor cancer evolution, evaluate treatment efficacy, and detect resistant subclones. These (longitudinal) multi-sample data allow the inference of phylogenetic trees, which provide intuitive diagrams and quantitative estimations to suggest hypotheses about evolutionary history. Phylogenetic inference is indispensable in studying species evolution and disease spread. They are increasingly used in studying cancer evolution and offer insights into intra-tumour heterogeneity, metastasis, and therapeutic resistance. We recently published a method, CNETML, to infer CNA-based trees from multiple samples of individual patients using a maximum likelihood approach, which was the first to allow the estimation of constant CNA rates in time from shallow whole genome sequencing data of longitudinal samples. We have applied CNETML to five patients with ovarian cancer to reveal their genomic histories. However, CNA rates are likely to change after certain events such as genome doubling. To allow estimating variable CNA rates in time, we will develop more realistic models of CNA evolution and better phylogenetic inference methods with wider applicability. To link the reconstructed tree with underlying mutational processes and aetiology, we will attach copy number signatures to nodes in the tree. To explore the landscape of CNA evolution, we will apply our methods to publicly available data of multi-region or longitudinal samples from 893 patients across 16 cancer types, including cell culture data. Similar to widely used phylogenetic inference programs in species evolution, our new programs will expand toolboxes for researchers in cancer evolution and promote applications of cancer phylogenetic inference. Our inferences will contribute to personalized therapy and public health. The estimated CNA rates may better distinguish the modes of copy number evolution and approximate chromosome mis-segregation rate, which are predictive of cancer progression or response to therapy. The estimated time of initiating CNAs indicates possible disease onset age, which may suggest potential time windows of screening and surveillance for cancers with unnoticeable early symptoms. The applications of our methods to data across cancer types will uncover similarities and differences in evolutionary trajectories among patients, which will contribute to the discovery of CNA drivers of tumour heterogeneity and patient stratification for personalized treatment. As our methods take CNAs as input, they may be generalized to other diseases featured by CNAs such as neurological disorders.