Project 466723

Cancer is driven by mutations that follow an evolutionary process, producing genetically distinct populations of cells known as clones. As a result, these clonal populations have varied responses to treatment; therefore, it is valuable to identify and characterize these clones when administering treatments in clinical applications. Liquid biopsies are samples of biological fluid (e.g. a blood sample) that contain DNA fragments known as cell-free DNA (cfDNA). Due to their non-invasive nature, they allow for serial sampling, enabling the study of clonal populations over time. In light of this, we propose a Bayesian statistical model to both estimate the relative proportions of existing clones, and discover emergences of novel clones. We will accomplish this task by analyzing the copy-number (CN) profiles of longitudinal cfDNA samples, while concurrently using the CN profile of a sequenced tissue biopsy taken at the beginning of treatment as an informative prior. Single-cell sequencing and whole-genome sequencing will be used for tissue biopsy and liquid biopsy respectively, whileMarkov Chain Monte Carlo and Variational Inference methods will be used to address intractable integrals common in Bayesian inference.Our model will elucidate the temporal heterogeneity of clonal populations over the course of treatment. Understanding the clonal landscape at several time points will enable practitioners to make informed treatment decisions to combat resistance, metastasis, or recurrence in cancer. This knowledge holds great utility, especially in personalized cancer care, as practitioners can track the responses of individual patients to therapies and make decisions on the fly - potentially improving the prognosis of hundreds of cancer patients at BC Cancer.