AbstractsBiology & Animal Science

The mapping task and its various applications in next-generation sequencing

by Christian Otto




Institution: Universität Leipzig
Department: Mathematik und Informatik
Degree: PhD
Year: 2015
Record ID: 1100786
Full text PDF: http://nbn-resolving.de/urn:nbn:de:bsz:15-qucosa-161623


Abstract

The aim of this thesis is the development and benchmarking of computational methods for the analysis of high-throughput data from tiling arrays and next-generation sequencing. Tiling arrays have been a mainstay of genome-wide transcriptomics, e.g., in the identification of functional elements in the human genome. Due to limitations of existing methods for the data analysis of this data, a novel statistical approach is presented that identifies expressed segments as significant differences from the background distribution and thus avoids dataset-specific parameters. This method detects differentially expressed segments in biological data with significantly lower false discovery rates and equivalent sensitivities compared to commonly used methods. In addition, it is also clearly superior in the recovery of exon-intron structures. Moreover, the search for local accumulations of expressed segments in tiling array data has led to the identification of very large expressed regions that may constitute a new class of macroRNAs. This thesis proceeds with next-generation sequencing for which various protocols have been devised to study genomic, transcriptomic, and epigenomic features. One of the first crucial steps in most NGS data analyses is the mapping of sequencing reads to a reference genome. This work introduces algorithmic methods to solve the mapping tasks for three major NGS protocols: DNA-seq, RNA-seq, and MethylC-seq. All methods have been thoroughly benchmarked and integrated into the segemehl mapping suite. First, mapping of DNA-seq data is facilitated by the core mapping algorithm of segemehl. Since the initial publication, it has been continuously updated and expanded. Here, extensive and reproducible benchmarks are presented that compare segemehl to state-of-the-art read aligners on various data sets. The results indicate that it is not only more sensitive in finding the optimal alignment with respect to the unit edit distance but also very specific compared to most commonly used alternative read mappers. These advantages are observable for both real and simulated reads, are largely independent of the read length and sequencing technology, but come at the cost of higher running time and memory consumption. Second, the split-read extension of segemehl, presented by Hoffmann, enables the mapping of RNA-seq data, a computationally more difficult form of the mapping task due to the occurrence of splicing. Here, the novel tool lack is presented, which aims to recover missed RNA-seq read alignments using de novo splice junction information. It performs very well in benchmarks and may thus be a beneficial extension to RNA-seq analysis pipelines. Third, a novel method is introduced that facilitates the mapping of bisulfite-treated sequencing data. This protocol is considered the gold standard in genome-wide studies of DNA methylation, one of the major epigenetic modifications in animals and plants. The treatment of DNA with sodium bisulfite selectively converts unmethylated cytosines to uracils, while methylated…