| A cell's transcriptome is characterized by the full repertoire of its condition-specific transcripts and their respective levels. Deciphering the transcriptome is essential for interpreting the functional elements of the genome, unraveling the molecular constituents of cells, and understanding disease. Furthermore, the emergence of next-generation DNA sequencing has significantly reduced costs, thereby revolutionizing the study of genomes and transcriptomes. These technologies have been leveraged for a number of applications, such as the sequencing of personal genomes on a large scale, which has revealed many novel variants and enabled the analysis of their effects on transcripts. In addition, as applied to transcriptome profiling (RNA-Seq), these technologies have allowed the study of transcripts at an unprecedented level. However, new computational methods are required to take advantage of the burgeoning volumes of data. In this thesis we present four computational approaches for transcript analysis in the context of next-generation DNA sequencing, including: (1) the Variant Annotation Tool, a computational framework to functionally annotate variants and assess their effects on the transcript structure of a gene; (2) RSEQtools, a modular approach for analyzing RNA-Seq data using compact anonymized data summaries; (3) FusionSeq, a tool for identifying fusion transcripts using paired-end RNA-Seq data; and (4) DupSeq, a computational approach for assessing the transcriptional activity of highly similar genomic sequences. Finally, as an application of these methods, we investigate the transcriptome dynamics of human embryonic stem cells as they differentiate into neural precursors. Together, these methodologies have been utilized extensively to gain novel insights into the transcriptome in different biological contexts. |
