A novel method for predicting protein phosphorylation via site-modification network profiles

Abstract

Protein phosphorylation, one of the most important types of post-translational modifications (PTMs), participates in multiple cellular processes. Accurate prediction on phosphorylaiton sites has become necessary, as many modifications are related to diseases and used as biomarkers. Currently a number of computational approaches only establish prediction models on sequence information. In this study, site-modification network (SMNet) profiles are proposed to enhance the prediction performance, which reflect information among in situ PTMs. In addition, a two-step algorithm that incorporates SVM with feature selection is adopted. To further demonstrate the method, we compare it with PPSP and GPS 3. 0, finally the results indicate that SMNet profiles effectively improve the performance on predicting phosphorylation sites.