Bioinformatics methods and approaches to discover disease variants from DNA sequencing data

Abstract

Next-generation sequencing is increasingly used to diagnose patients with suspected genetic disease. Yet, even after exome or whole genome sequencing, many patients remain undiagnosed. In many cases a genetic diagnosis is not made because we either failed to detect the causal variant, or succeeded in detecting it, but failed to identify it as causative. There is a clear need to develop novel bioinformatics methods and sequencing strategies to address these shortcomings and to increase diagnostic rates. In this thesis I develop several strategies to address these issues. I propose a pooled-parent exome sequencing approach to prioritise de novo variants for genetic disease diagnosis. In this strategy, a set of probands have individual exome sequencing, while the DNA from all the parents of the probands are pooled, exome captured and sequenced together. The variants called in this pool are used to filter out inherited variants in the probands so the remaining list is enriched for de novo variants. Short Tandem Repeat (STR) expansions are a class of disease-causing variants that are frequently missed in short read sequencing data. Here I develop and validate STRetch, a new bioinformatics method to detect STR expansions using STR decoy chromosomes. I show that STRetch can be used to detect both known pathogenic STR expansions, and novel expansions at other annotated STR loci across the genome. I further use STRetch to explore variation across hundreds of individuals to inform our understanding of what is common variation and what is potentially pathogenic, to aid in prioritising STR variants in a gene-discovery setting. Some of the methods that I have developed and describe within this thesis have already been used to help patients receive a genetic diagnosis.