Identification and Functional Annotation of Alternatively Spliced Isoforms

Abstract

Alternative splicing is a key mechanism for increasing the complexity of transcriptome and proteome in eukaryotic cells. A large portion of multi-exon genes in humans undergo alternative splicing, and this can have significant functional consequences as the proteins translated from alternatively spliced mRNA might have different amino acid sequences and structures. The study of alternative splicing events has been accelerated by the next-generation sequencing technology. However, reconstruction of transcripts from short-read RNA sequencing is not sufficiently accurate. Recent progress in single-molecule long-read sequencing has provided researchers alternative ways to help solve this problem. With the help of both short and long RNA sequencing technologies, tens of thousands of splice isoforms have been catalogued in humans and other species, but relatively few of the protein products of splice isoforms have been characterized functionally, structurally and biochemically. The scope of this dissertation includes using short and long RNA sequencing reads together for the purpose of transcript reconstruction, and using high-throughput RNA-sequencing data and gene ontology functional annotations on gene level to predict functions for alternatively spliced isoforms in mouse and human. In the first chapter, I give an introduction of alternative splicing and discuss the existing studies where next generation sequencing is used for transcript identification. Then, I define the isoform function prediction problem, and explain how it differs from better known gene function prediction problem. In the second chapter of this dissertation, I describe our study where the overall transcriptome of kidney is studied using both long reads from PacBio platform and RNA-seq short reads from Illumina platform. We used short reads to validate full-length transcripts found by long PacBio reads, and generated two high quality sets of transcript isoforms that are expressed in glomerular and tubulointerstitial compartments. In the third chapter, I describe our generic framework, where we implemented and evaluated several related algorithms for isoform function prediction for mouse isoforms. We tested these algorithms through both computational evaluation and experimental validation of the predicted ‘responsible’ isoform(s) and the predicted disparate functions of the isoforms of Cdkn2a and of Anxa6. Our algorithm is the first effort to predict and differentiate isoform functions through large-scale genomic data integration. In the fourth chapter, I present the extension of isoform function prediction study to the protein coding isoforms in human. We used a similar multiple instance learning (MIL)-based approach for predicting the function of protein coding splice variants in human. We evaluated our predictions using literature evidence of ADAM15, LMNA/C, and DMXL2 genes. And in the fifth and final chapter, I give a summary of previous chapters and outline the future directions for alternatively spliced isoform reconstruction and function prediction studies.