Bioinformatic challenges in translating high throughput sequencing into clinical settings

© 2016 Dr. Simon Paul SadedinIn the last ten years, genomic sequencing has undergone a technological revolution, enabling large numbers of genes, or even entire exomes or genomes to be sequenced economically. These capabilities, based around the new technology of high throughput sequencing (HTS), have had a transformational impact on biomedical research. The impact in clinical settings is likely to be equally as profound due to the potential to enhance the diagnosis, treatment and management of disease. These impacts, however, are only beginning to be explored. Clinical genetics is in a transitional phase, as the new technologies are being validated and tailored to the clinical setting. In that process, significant new challenges are being faced. This thesis focuses on one particular aspect: the essential role of bioinformatics in the translational process. The thesis discusses the bioinformatics of clinical translation from a variety of points of view. In the first chapter, an overview of the current clinical setting is presented. This includes a description of the general challenges of clinical translation, background about the history and context of modern clinical genetics, and the key bioinformatic methods that are applied to derive clinical results. The second chapter consists of a published paper describing an analysis pipeline, Cpipe. This analysis pipeline was developed as part of the Melbourne Genomics Health Alliance demonstration project, a prototype for deployment of clinical exome sequencing. It illustrates many of the key issues associated with moving high throughput sequencing into a clinical setting. In the third chapter, a specialised sequencing technology, HaloPlex, is described in detail. HaloPlex is in use at the Murdoch Childrens Research Institute as part of a sequencing project that aims to diagnose disorders of sexual development. HaloPlex data has many unique characteristics, and these form the basis of several subsequent chapters of the thesis. The fourth chapter extends the discussion of HaloPlex into an evaluation of the accuracy of variant detection from HaloPlex data. The key problems in analysing HaloPlex data are identified, and one particular issue, adapter contamination, is discussed in depth. A novel method is described for removal of adapter contamination in HaloPlex data. The fifth chapter discusses the difficulty of detecting some clinically relevant types of mutation from HTS data. Exon and larger size deletions are identified as particularly important. The chapter introduces a novel simulation method, Ximmer, which is specifically designed to simulate deletions in targeted sequencing data. Using the simulation method, we present an evaluation of performance of existing deletion detection methods. Lastly, the thesis returns to HaloPlex to consider a dedicated HaloPlex algorithm for identifying single copy deletions. An evaluation of current methods is presented and the statistical basis of a method, called Angel, is described and evaluated to show that significantly improved performance is attained by leveraging the unique characteristics the HaloPlex technology. In summary, this thesis describes the bioinformatic context of clinical genomics, and details several related novel methods to address key challenges that are identified. These include a clinical exome sequencing pipeline, a method for simulating deletions in exome sequencing data, and a method for detection of deletions in HaloPlex sequencing data