SuccSPred: Succinylation Sites Prediction Using Fused Feature Representation and Ranking Method

Abstract

Protein succinylation is a novel type of post-translational modification in recent decade years. Experiments verified that it played an important role in biological structure and functions. However, experimental identification of succinylation sites is time-consuming and laborious. Traditional technology cannot meet the rapid growth of the sequence data sets. Therefore, we proposed a new computational method named SuccSPred to predict succinylation sites in a given protein sequence by fusing many kinds of feature representation and ranking method. SuccSPred was implemented based on a two-step strategy. Firstly, linear discriminant analysis was used to reduce feature dimensions to prevent overfitting. Subsequently, the predictor was built based on incrementing features selection binding classifiers to identify succinylation sites. After the comparison of the classifiers using ten-fold cross-validation experiment, the selected model achieved promising improvement. Comparative experiments showed that SuccSPred significantly outperformed previous tools and had the great ability to identify the succinylation sites in given proteins.