O-glycosylation Site Prediction Using Randome Forest Importance and Support Vector Machine

Due to the significant roles of O-glycosylation in the biological processes as well as its association with various human diseases, in this work we propose a machine learning model to improve O-glycosylation site prediction from the protein sequences. Different types of sequence-based feature extraction methods were used to encode the peptide sequences. Consequently, random forest importance and incremental feature selection techniques are implemented to select small informative features. Then, the support vector machine (SVM) method was used for classification and prediction. In terms of accuracy, AUC, sensitivity, specificity, and MCC metrics, the proposed model outperformed the other machine learning techniques and the existing tools.