A two-stage method for O-glycosylation site prediction

Correctly predicting the site of O-glycosylation will greatly benefit the search and design of new specific and efficient GalNAc-transferase inhibitors. In this article, the site of O-glycosylation was studied using the correlation-based feature subset (CfsSubset) selection method combined with a wrapper method. Twenty-three important biochemical features were found based on a jackknife test from original data set containing 4779 features. By using the AdaBoost method with the twenty-three selected features, the prediction model yields an accuracy rate of 88.1% for the jackknife test and 87.5% for an independent set test, with increased accuracy over the original dataset by 8.5% and 10.42%, respectively. It is expected that our feature selection scheme can be referred to as a useful assistant technique for finding effective competitive inhibitors of GalNAc-transferase. An online predictor based on this research is available at http://chemdata.shu.edu.cn/gal_p/.

► Fewer features (23 features) were selected from original data set (4779) features.
► Higher prediction accuracies obtain for jackknife test (88.1%) and independent set test (87.5%).
► Subsites P3, P1′ and P3′ are closely related to O-glycosylation.
► Secondary structure of amino acid residues shows high correlation to to O-glycosylation.