| A majority of eukaryotic genes have alternating exons and introns. Introns in pre-mRNAs are removed and exons are joined to generate mature transcripts in a process called splicing. By different combinations of exons and splice sites, i.e., alternative splicing, one gene can produce multiple transcript and protein isoforms, providing a major source of proteomic diversity, novel mechanisms of gene expression regulation, and new paths of gene evolution. Splicing and alternative splicing are dictated by interactions of many cis-regulatory elements and trans-acting splicing factors in a cellular machinery called spliceosome. However, the splicing code that elucidates how these interactions determine the splicing outcome, sometimes specific for particular tissues, developmental stages, and different species or populations, is still poorly understood. This study aims to advance the mechanistic understanding of the fidelity and regulation of both constitutive and alternative splicing. To approach the aim, I mainly use statistical and computational analysis of genomewide, high-throughput data to generate experimentally testable hypotheses, combined with experimental validations by collaborative bench biologists. In this dissertation, I first demonstrate the limited fidelity of splicing and describe a new and unusual type of alternative splice site---dual-specificity splice site, which implies the evolutionary selective pressure to reduce splicing fluctuations and eliminate evolutionary intermediates. Then, I show how such selective pressure results in non-random, distinct distributions of splicing-regulatory elements of different classes in exons and introns, including deep intronic sequences, for optimal exon and intron discrimination. The distribution of splicing-regulatory elements is gauged by a neutral evolution model developed from DNA strand-asymmetry patterns, and the principle is also very effective to predict new regulatory elements. To achieve a better understanding of the organization and functional impacts of splicing-regulatory networks, I use the tissue-specific splicing factors Fox-1/2 as a model. By comparative analysis of 28 vertebrate species, this study predicts thousands of conserved Fox-1/2 targets with high specificity and sensitivity; at least 50-60% of predicted targets can be experimentally verified in HeLa cells. This analysis reveals a surprising extensiveness and complex patterns of tissue-specific splicing regulation. The regulatory network is highly organized and modular, with many predicted targets important for neuromuscular functions and diseases. Lastly, in addition to splicing fidelity and regulation in normal conditions, I also describe one of the first surveys of splicing dysregulation in prostate cancer, which demonstrates the importance of splicing in tumorigenesis, and the unique advantage of splicing profiling in cancer sample classification and biomarker identification. |
