A Combined SVM and Markov Model approach for Splice Site Identification

Abstract

Due to an exponential increase in biological sequence data, gene detection has become one of the challenging tasks in computational biology. Splice site prediction is an essential part of the gene detection. Thus, it has great significance to develop efficient methods for accurately identifying splice sites. This paper introduces a novel algorithm to predict the splice sites based on support vector machine (SVM) and a new type of Markov chain model, namely DMM2. The proposed method shows great improvement over most of the current state of art methods, including MM1-SVM, Reduced MM1-SVM, SVM-B, LVMM, MMI-RF, MM2F-SVM, MCM-SVM, DM-SVM and DM2-AdaBoost. The repeated 10-fold cross validation was used to assess the performance of the method on the HS3D dataset. In addition, we applied it to NN269 dataset to examine the stability of the proposed method. The experimental results indicate that the new approach is feasible and efficient.