An effective statistical moment-based feature extraction technique to identify the phosphoglycerylation sites from protein sequences

A kind of covalent modification known as post-translational modification (PTM) happens following the biosynthesis process, which is important in cell biology research. A reversible PTM called Lysine phosphoglycerylation alters glycolytic enzyme activity and is linked to several disorders, including heart failure, arthritis, and nervous system deterioration. Identification of phosphoglycerylation has been improved using a variety of feature extraction approaches with machine learning technologies. However, it may still be improved by developing new and more powerful feature extraction algorithms and classification approaches. In this study, an efficient feature extraction technique named statistical moment of physicochemical properties (SMPP) is suggested using the statistical moment procedure and physicochemical characteristics of amino acids for generating numerical features from proteins. Next, a computational model is created with a support vector machine (SVM) to measure the predictive capability of SMPP in the identification of phosphoglycerylation sites. An overall average accuracy of 98.26 % on 10-fold cross-validation and 99.40 % on the independent test is acquired by SMPP features, which is better than that of the experimental results of currently available feature extraction techniques used for identifying the phosphoglycerylation sites. The web server, dataset, and source code for this research are openly obtainable at https://shaikot.pythonanywhere.com/smfeature/.