STRUCTURAL VARIATION CALLING IN HIGH-THROUGHPUT NEXT-GENERATION SEQUENCING DATA

Abstract

Structural Variations (SVs) are large-scale mutations in the genome. They are accountable for a large portion of heritable differences between individuals, and they have been proven to play a key role in phenotypic variability and in devastating genetic diseases. For more than a decade methods predicting SVs from second generation sequencing datasets have been developed, but when tested on biological datasets their performance is generally poor. In this thesis, we tackle the problem of predicting several major classes of SVs from DNA sequencing data. For each class we show the shortcomings of the existing solutions, propose technique to address them, and prove on both simulated and real biological datasets that our solutions outperform the state-of-the-art methods in SV calling.