SVM-based prediction of protein-protein interactions of Glucosinolate biosynthesis

Abstract

Protein-protein interactions (PPIs) are of biological interest because they orchestrate a number of cellular processes such as metabolic pathways and immunological recognition. This paper aims at exploring more PPIs of Glucosinolates biosynthetic pathways and removing PPIs falsely predicted. A support vector machine (SVM) predictor with the radial basis kernel function (RBF kernel) is trained based on the domain and domain-domain interaction (DDI) information of the amino acid sequences. In this paper, a symmetrical pair of feature vectors is used to represent the symmetrical relationship between two proteins, and 5-fold cross-validation is used to search the best SVM parameters. Then the best SVM parameters are used to train the SVM-based PPIs predictor. The proteins originate from gene AT4G14800 and ATSGS4810 (ID of Arabidopsis Genome Initiative (AGI)), ATSGOS730 and AT4G18040, ATlG04S10 and ATSGOS260 are affirmed interactive by this SVM-based PPIs predictor.