Ultra-large-scale genomics approaches to improve cancer therapeutic response

Abstract

The advent of low-cost, high-throughput sequencing technologies has elucidated targetable cancer-specific genomic alterations and allowed the development of precision therapies and their deployment into clinical use. However, definitive determinants of response to targeted drugs remain elusive. Therefore, knowledge of genomic features relevant to mechanisms of oncogenesis and cancer susceptibility to therapeutic interception is imperative. In this dissertation, I detail studies that identify such features, through the analysis of publicly available RNA sequencing (RNA-seq) datasets at a massive scale. First, we performed some of the largest RNA-seq analyses ever conducted to identify branchpoint nucleotide positions genome-wide, characterize the unexpected structural and regulatory complexity of human introns, and describe the unappreciated genome-wide prevalence of circular intron-derived RNAs. Second, we coupled multiple large genetic screens with experimental perturbations and RNA-seq analyses of patient-derived and cell line models of acute myeloid leukemia (AML) to identify determinants of drug response. We determined that splicing modulation is a unique AML susceptibility and identified specific splicing changes in the transcripts of spliceosomal components and apoptotic factors mediating sensitivity and resistance to the BCL2 inhibitor venetoclax. Last, we analyzed large and diverse cancer cohorts using genomic, statistical, and machine learning methods to identify DUX4 reactivation as a common mechanism of immune evasion in advanced cancers and central feature of metastatic cancer patients resistant to checkpoint immunotherapy. These works amalgamate genomic and clinical data to identify novel strategies to target cancer and improve the precision and efficacy of current treatment modalities.