UC San Diego

Segmental duplications or low-copy repeats (LCRs) are long segments of duplicated DNA that cover more than 5% of the human genome and overlap more than 600 protein-coding genes. Copy number and sequence variants in over 150 such duplicated genes (e.g. SMN1/2, STRC, NCF1) are associated with risk for rare and complex human diseases. Paralogous sequence variants (PSVs) are short differences between homologous sequences within duplicated loci. It has been shown that many PSVs are not fixed in the population, which reduces their potential to differentiate paralogous regions. Moreover, segmental duplications exhibit extensive copy number variation, and are characterized by poor read mappability even for long-read data. All these factors lead to diminished accuracy of existing bioinformatical tools for short- and long-read data in duplicated regions. This dissertation presents three novel computational methods that solve classical bioinformatical problems (read mapping, variant calling and copy number variation detection) in LCR regions. In contrast to existing tools, three proposed methods examine PSV genotypes in order to distinguish sets of reliable and unreliable PSVs, and use reliable PSVs to achieve higher accuracy than state-of-the-art methods in the field.

First, we describe a probabilistic method, DuploMap, designed to improve the accuracy of long-read mapping within LCR regions. It iteratively genotypes PSVs and leverages reliable PSVs to distinguish between candidate read locations. This allows for high accuracy variant calling in segmental duplications using long reads. Next, we present the first toolkit for LCR regions, Parascopy. Parascopy uses short-read whole-genome sequencing to estimate total copy number as well as paralog-specific copy number for duplicated genes. Parascopy analyzes reads mapped to different repeat copies and utilizes multiple samples to mitigate sequencing bias and identify reliable PSVs. Accurate copy number estimation facilitates discovery of pathogenic copy number changes in duplicated genes. A novel variant caller, ParascopyVC, builds upon copy number variation detection and uses short-read data to call pooled and locus-specific variants within segmental duplications. ParascopyVC uses population allele frequencies and pooled genotypes to select informative PSVs. Finally, the tool uses informative PSVs to identify additional locus-specific variants, enabling the discovery of novel disease-causing variants in duplicated genes.