iDPGK: characterization and identification of lysine phosphoglycerylation sites based on sequence-based features

Area of Science:

• Biochemistry and Molecular Biology
• Bioinformatics and Computational Biology

Background:

• Protein phosphoglycerylation is a non-enzymatic post-translational modification involving 1,3-bisphosphoglyceric acid (1,3-BPG) and lysine residues.
• This modification plays a regulatory role in glucose metabolism and the glycolytic process.
• Limited resources exist for computational identification of phosphoglycerylation sites, necessitating advanced prediction methods.

Purpose of the Study:

• To investigate the sequence-based characteristics of protein phosphoglycerylation sites.
• To develop a computational model for predicting phosphoglycerylation sites.

Main Methods:

• Bioinformatics analysis using TwoSampleLogo and PTM-Logo to identify sequence features.
• Feature selection based on F-score ranking.
• Development of prediction models using Decision Trees (DT), Random Forests (RF), and Support Vector Machines (SVM) classifiers.
• Evaluation using five-fold cross-validation and an independent testing set.

Main Results:

• Positively charged amino acids are enriched in regions surrounding phosphoglycerylation sites, particularly upstream.
• Non-polar and aliphatic amino acids are also abundant near modified lysines, aiding discrimination.
• The SVM model achieved high performance, with 77.5% sensitivity and 73.6% specificity on the training set.
• The model demonstrated consistent performance on an independent test set (75.7% sensitivity, 64.9% specificity).

Conclusions:

• Sequence-based features are effective for predicting protein phosphoglycerylation sites.
• The developed SVM model, implemented as the iDPGK web server, provides a valuable tool for researchers.
• iDPGK is freely available for computational identification of phosphoglycerylation sites.