Accurate Prediction of Alternatively Spliced Cassette Exons Using Evolutionary Conservation Information and Logitlinear Model

Accurate prediction of alternative splicing (AS) is important for understanding the mechanism of gene regulation and studying the pathogenesis of diseases. By analyzing the evolutionary conservation information in alternatively spliced cassette exons (SCEs), constitutively spliced exons (CSEs) and their flanking intronic regions, we found that the average conservation score (AConScore) distributions of SCEs are significantly different from those of CSEs. We further integrated these unique features into the logitlinear model, and constructed a classifier, named AEDetector, for distinguishing SCEs from CSEs. The five-fold cross validation results show that the sensitivity and specificity of AEDetector are higher than those recently reported on the same dataset.