Computational Methods for Integrative Structural Variant Analysis Across Species Boundaries

Abstract

Structural variations (SVs) are a phenomenon that have a tremendous impact on all species. SVs are the result of fundamental rearrangement mechanisms but can lead to severe human diseases like cancer. Rearrangement events also provide means that enable bacteria to adapt to environmental pressures where they can also happen across species boundaries in events called horizontal gene transfer (HGT). The incorporation of foreign genes from a donor into an acceptor genome can be investigated on the genomic level, the activity and protein expression changes, however, are better revealed on the proteomic level. This thesis contributes four computational methods for the detection of complex SVs of various types and sizes including HGT events from genomic next-generation sequencing (NGS) data and proteomic shotgun mass-spectrometry (MS) data. Concerning HGT events, our methods address the questions of what organisms are involved in the transfer, what genes are exactly transferred and to what position, and what are the implications on proteomic level. First, we present the generic SV detection tool Gustaf. Gustaf improves the size and type resolution compared to previous SV detection methods. A further specific advantage is the characterisation of translocations and dispersed duplications as a combination of simple, delocalised variants that have to be inferred from separate SV calls. With this basis for a more in-depth focus on HGT detection, we developed two mapping-based methods, Daisy and DaisyGPS. Daisy facilitates Gustaf and further SV detection strategies to precisely identify the transferred region within the donor and its insertion site in the acceptor genome. DaisyGPS uses metagenomic profiling strategies to identify suitable acceptor and donor references. In contrast to previous approaches based on sequence composition patterns or phylogenetic disagreements, our methods provide a detection based on sequence comparison and hence offer novel means of evidence. In the last project, we present a method for HGT detection, called Hortense, that is based on proteomic MS data. Hortense extends a standard database peptide search with a thorough cross-validation to ensure HGT properties, and is the first dedicated proteomics HGT detection method. Results from Hortense can also serve as supporting evidence and functional confirmation for HGT events proposed by our genomic-based methods. Taken together, the three HGT methods provide a full view of the transfer event that was not be possible before or with one of the methods alone.