The Identification of Human Cryptic Exons Based on SVM

Alternative splicing is the main mechanism expanding proteome diversity. Cryptic exon is an important alternative splicing form that is different from skipped exon in statistical characters. We introduce a process to identify the human cryptic exons in candidate ones. This method contains two steps performed by two classifiers based on the support vector machine (SVM). The first classifier distinguishes authentic exons from pseudo exons; the second classifier distinguishes cryptic exons from constitutive and skipped exons. It can achieve the accuracy of 94.25% and 69.75% in the two steps, respectively. This method uses no expressed sequence tags (ESTs) or conservation information, so it can be used more widely and easily. The performances are higher than an existing method predicting splice sites of cryptic exons.