PROJECT SUMMARY Complex structural variants (SVs) are a class of mutations consisting of clustered DNA copy number changes and rearrangements. These alterations can produce driver mutations in cancer but have been underexplored due to the analytical limitations of bulk short read sequencing. There is an urgent need for new computational tools that can apply long-range and single cell genomic profiles to structural variant analysis. Genome graphs are a computational framework that can be extended to these new data modalities to study the allelic structure and evolution of complex SVs. Recent genome graph analysis of pan-cancer whole genomes by our lab identified a novel complex structural variant pattern termed pyrgo. Pyrgo consist of “towers” of clustered tandem duplications and are enriched in prostate and ovarian adenocarcinomas. In Aim 1, we will construct haplotype graphs to characterize the allelic structure of pyrgo. Haplotype graphs represent allele-specific genomic segment and junction copy numbers and will be inferred from long range profiling data. Using these graphs, we will characterize the parental and somatic allele structure of pyrgo duplications in pan-cancer genomes. We will identify associations between allelic structure and cell-of-origin, prior systemic therapy, and genomic features such as chromatin loops and replication timing. In Aim 2, we will construct single cell genome graphs that recapitulate the evolution of pyrgo. Single cell genome graphs are a set of phylogenetically linked genome graphs representing the structural variants present in individual cells, along with the ancestral subclone in which each aberrant genomic junction first arose. These graphs will be inferred from single cell whole genome profiles. We will use single cell genome graphs to model the acquisition of tandem duplications comprising pyrgo during tumor evolution in ovarian adenocarcinoma tissue samples.

PROJECT NARRATIVE Complex structural variants are a frequent but underexplored aspect of cancer genomes that may drive aggressive tumors and reflect treatable defects in DNA repair. While long-range and single cell genomic profiling technologies have matured, algorithms for inferring complex structural variants from these data are lacking. This proposal will develop such algorithms and apply them to study the evolution of pyrgo, a recently identified complex structural variant class prevalent in ovarian and prostate cancer.