Tissue-specific classification of alternatively spliced human exons

Abstract

Alternative splicing is involved in numerous cellular functions and is often disrupted and involved in disease. Previous research has identified methods to distinguish alternative conserved exons (ACEs) in human and mouse. However, the cellular machinery, the spliceosome, does not use comparative genomics to decide when to include and when to exclude an exon. Human RefSeq exons obtained from the University of California Santa Cruz (UCSC) genome browser were analyzed for tissue-specific skipping. Expressed sequence tags (ESTs) were aligned to exons and their tissue of origin and histology were identified. ACEs were also identified as a subset of the skipped exons. About 18% of the exons were identified as tissue-specifically skipped in one of sixteen different tissues at four stringency levels. The different datasets were analyzed for both general features such as exon and intron length, splice site strength, base composition, conservation, modularity, and susceptibility to nonsense-mediated mRNA decay caused by skipping. Cis-element motifs that might bind protein factors that affect splicing were identified using overrepresentation analysis and conserved occurrence rate between human and mouse.

(cont.) Tissue-specific skipped exons were then classified with both a decision-tree based classifier (Random ForestsTM) and a support vector machine. Classification results were better for tissue-specific skipped exons vs. constitutive exons than for tissue-specific skipped exons vs. exons skipped in other tissues.