UC San Diego

Genomic instability is a hallmark of cancer and one of the key forces driving tumor evolution. It can arise from intrinsic factors, such as defects in DNA repair pathways, and extrinsic influences, including the microbiome residing within the tumor and its microenvironment. Understanding these processes is essential for elucidating tumor initiation and evolution. In this dissertation, I investigated how intrinsic genomic defects and extrinsic microbial factors contribute to genomic instability across multiple cancer types. In Chapter 2, I introduce HRProfiler, a machine learning tool developed to detect homologous recombination deficiency, a DNA repair defect that increases sensitivity to platinum-based chemotherapy and PARP inhibitors. HRProfiler leverages six mutational features from whole-genome or whole-exome sequencing of breast and ovarian cancers, outperforming existing methods. In retrospective analyses of clinical trials and real-world datasets, HRProfiler accurately predicts responses to PARP inhibitors and platinum therapies, highlighting its potential as a clinically actionable biomarker. In Chapter 3, I explored the role of Epstein-Barr virus (EBV) in promoting genomic instability in EBV-associated cancers. Our experiments showed that the viral protein EBNA1 binds specifically to a repetitive region on chromosome 11q23, inducing site-specific DNA breakage in the human genome. Analysis of EBV-associated cancer genomes reveals a significant enrichment of structural variants on chromosome 11, suggesting that EBV-induced chromosomal fragility contributes to cancer development. In Chapter 4, I explored geographic variations in the microbiome of 963 colorectal cancers (CRC) from 11 countries. I first developed a robust microbiome detection pipeline based on best practices and implemented it using Nextflow for scalability and reproducibility. I then applied this pipeline to the CRCs and observed a strong positive correlation between microbial diversity and cancer incidence rates. My analysis also revealed significant associations between specific bacterial genera and mutations in key CRC driver genes. Finally, using a novel nonnegative matrix factorization approach, I identified distinct microbial communities enriched in CRC. Together, this work deepens our understanding of how intrinsic genomic alterations and the tumor microbiome can drive genomic instability. These insights have direct implications for improving cancer diagnostics, therapeutic targeting, and risk stratification across multiple cancer types.