Novel graph based algorithms for transcriptome sequence analysis

Abstract

RNA-sequencing (RNA-seq) is one of the most-widely used techniques in molecular biology. A key bioinformatics task in any RNA-seq workflow is the assembling the reads. As the size of transcriptomics data sets is constantly increasing, scalable and accurate assembly approaches have to be developed.Here, we propose several approaches to improve assembling of RNA-seq data generated by second-generation sequencing technologies. We demonstrated that the systematic removal of irrelevant reads from a high coverage dataset prior to assembly, reduces runtime and improves the quality of the assembly. Further, we propose a novel RNA-seq assembly work- flow comprised of read error correction, normalization, assembly with informed parameter selection and transcript-level expression computation. In recent years, the popularity of third-generation sequencing technologies in- creased as long reads allow for accurate isoform quantification and gene-fusion detection, which is essential for biomedical research. We present a sequence-to-graph alignment method to detect and to quantify transcripts for third-generation sequencing data. Also, we propose the first gene-fusion prediction tool which is specifically tailored towards long-read data and hence achieves accurate expression estimation even on complex data sets. Moreover, our method predicted experimentally verified fusion events along with some novel events, which can be validated in the future.