iDPGK

iDPGK predicts lysine phosphoglycerylation sites in protein sequences, identifying lysine residues modified by 1,3-bisphosphoglyceric acid (1,3-BPG) to form 3-phosphoglyceryl-lysine (pgK), to facilitate study of this post-translational modification in glucose metabolism and glycolysis.

Key Features:

• Sequence-Based Feature Analysis: Performs TwoSampleLogo and PTM-Logo analyses showing enrichment of positively charged amino acids in upstream flanking regions and increased prevalence of non-polar and aliphatic residues around phosphoglycerylated lysines.
• Comprehensive Feature Set: Incorporates amino acid composition, amino acid pair composition, positional weighted matrix, and position-specific scoring matrix (PSSM) as sequence-derived features.
• Advanced Machine Learning Models: Evaluates Decision Trees (DT), Random Forests (RF), and Support Vector Machines (SVM) with feature ranking by F-score to select discriminative features.
• Robust Validation: An SVM trained on selected sequence-based features achieved sensitivity 77.5%, specificity 73.6%, accuracy 74.9%, and Matthews Correlation Coefficient (MCC) 0.49 in five-fold cross-validation and maintained sensitivity 75.7% and specificity 64.9% on an independent test set.

Scientific Applications:

• Phosphoglycerylation site mapping: Enables computational identification of 3-phosphoglyceryl-lysine (pgK) sites to guide experimental validation.
• Metabolic regulation studies: Supports investigation of phosphoglycerylation's regulatory roles in glucose metabolism and glycolysis.
• Complement to experimental workflows: Provides predictions useful when experimental verification of phosphoglycerylation sites is limited or challenging.

Methodology:

Sequence-based feature extraction (amino acid composition, amino acid pair composition, positional weighted matrix, PSSM), TwoSampleLogo and PTM-Logo analyses, F-score feature ranking, classification using DT, RF and SVM, and validation by five-fold cross-validation and independent testing.