Structural variations: detection and annotation in cancer genomes

Abstract

Structural variations (SVs) are genomic variants that typically impact more than 50 nucleotides in length and significantly contribute to cancer development and evolution. However, it is challenging to accurately infer and classify SVs in full range and type using short-read next-generation sequencing (NGS) technologies, which will also limit downstream annotation efforts to understanding their oncogenic impact. This PhD thesis addresses the challenges of SV detection and annotation in cancer genomics. Firstly, Chapter 1 summarises current methods and limitations for inferring somatic SVs. A comprehensive evaluation study is conducted in Chapter 2 to assess the extent to which various common factors impact SV detection accuracy, and hence should be considered in whole-genome sequencing (WGS) study designs. Shiny-SoSV, a web-based interactive performance calculator is developed in Chapter 3 for estimation and comparison of somatic SV detection sensitivity and precision based on any combinations of user-definable parameters. In Chapter 4, a “real-life” application of somatic SV detection and annotation is conducted for a large-scale prostate cancer genomics study, providing insights into the role of SV in cancer development. Finally, in Chapter 5, two validation approaches, using visual inspection and long-read sequencing data, are evaluated and compared to provide guidance on a “best practise” for SV validation. Overall, this PhD work highlights, and offers solutions to overcome challenges associated with SV detection and annotation, and illustrates the power of incorporating SVs in cancer genomic studies.