Performance evaluation of feature extraction to improve the classification of PTM in C-glycosylation using XGBoost

Abstract

Protein function is regulated by an important mechanism known as post-translational modification (PTM). Covalent and enzymatic protein modifications are added during protein biosynthesis, and such alterations significantly influence the regulation of gene activity and the functionality of proteins. Glycosylation, one type of PTM, involves adding sugar groups to a protein's structure. Numerous illnesses, such as diabetes, cancer, and the flu, have been linked to glycosylation. Therefore, it is critical to predict the presence of glycosylation, whether it occurs or not. Currently, predicting glycosylation sites is still done manually using biological methods, which require repeated experiments and a significant amount of time. To address these challenges, it is essential to rapidly develop computational data models using machine learning methods. In this study, the extreme gradient boosting (XGBoost) method is implemented, and C-glycosylation data is obtained from the publicly accessible UniProt website. The objective is to enhance the accuracy of C-glycosylation prediction using the XGBoost method. Feature extraction is performed using amino acid index (AAindex), composition, transition, and distribution (CTD), solvent AccessiBiLitiEs (SABLE), hydrophobicity, and pseudo amino acid composition (PseAAC) to improve accuracy. The minimum redundancy maximum relevance (MRMR) method is applied for feature selection. The findings of the study demonstrate that the PTM C-glycosylation prediction achieved 100%.