On Reconstructing Single-Cell Cancer Evolutionary Histories

Cancer is the result of a complex evolutionary process during which cells accumulate genetic and non-genetic alterations over time. Investigating the evolutionary patterns of these alterations is key to revealing mechanisms that drive disease progression and to improving therapeutic approaches. Most previous cancer evolutionary studies have relied on bulk sequencing, which averages signal from thousands of cells and is thus primarily powered to detect early, high-frequency alterations. Yet small subpopulations, or clones, of cells often play a key role in progression, indicating that later, low-frequency alterations can be mechanistically important. In this thesis, I demonstrate how recent, scalable single-cell whole-genome DNA sequencing (scDNA-seq) overcomes these limitations and enables novel insights into cancer evolution. I present a novel primary–metastasis–matched non-small-cell lung cancer patient scDNA-seq dataset containing thousands of individual cells. To analyse the evolutionary patterns in this dataset, I introduce a statistical framework to infer the evolution of both genetic and non-genetic alterations in single-cell cancer clones. I apply this framework to reveal that the metastatic potential of certain clones may be associated with non-genetic rather than genetic alterations. This demonstrates the value of single-cell clone evolutionary analysis, yet it does not resolve which alterations occur together within individual cells or in the smallest clones. Therefore, to fully exploit native single-cell resolution, I developed PHANTOM, a novel statistical algorithm that reconstructs single-cell evolutionary histories. Through extensive benchmarking, I demonstrate that PHANTOM enables evolutionary reconstruction with high accuracy from scDNA-seq data, in contrast to previous state-of-the-art approaches. Applying PHANTOM to a large cohort of more than 60,000 cells across multiple tumour types, I reveal that both genetic and non-genetic factors, including DNA-repair defects, reshape cancer evolution. Together, these datasets, algorithms, and analyses advance our understanding of single-cell cancer evolutionary histories.