glycopp

glycopp predicts N- and O-glycosylation sites in prokaryotic protein sequences to identify asparagine-linked (N-linked) and serine/threonine/tyrosine-linked (O-linked) glycosites for prokaryotic glycoproteome analysis.

Key Features:

• SVM-based models: Uses Support Vector Machine classifiers developed specifically for predicting glycosylated residues (glycosites).
• Feature representation: Employs binary profiles, composition profiles, and position-specific scoring matrix (PSSM) profiles as training features.
• Hybrid modeling: Integrates predicted secondary structure and accessible surface area with sequence-derived features to account for glycosylation on folded proteins.
• Training dataset: Trained on 107 experimentally validated N-linked and 116 O-linked glycosites extracted from 59 prokaryotic glycoproteins.
• Taxonomic coverage: Includes N-glycosites from Crenarchaeota, Euryarchaeota and Proteobacteria, and O-glycosites from Actinobacteria, Bacteroidetes, Firmicutes, and Proteobacteria.
• Performance metrics: Reported N-glycosite accuracy 82.71% (MCC = 0.65) and O-glycosite accuracy 73.71% (MCC = 0.48).
• Independent validation: Best-performing models validated on a separate set of 28 prokaryotic glycoproteins.

Scientific Applications:

• Glycosite prediction: Identification of potential N- and O-glycosylation sites in prokaryotic proteins for downstream experimental design.
• Glycoproteome characterization: Exploration of glycosylation patterns across diverse prokaryotic phyla to support comparative glycoproteomics.
• Structure–function inference: Support for hypotheses relating protein glycosylation to structural accessibility and functional modulation.

Methodology:

SVM classifiers were trained using binary, composition and PSSM profiles derived from 107 N-linked and 116 O-linked glycosites (59 proteins), with hybrid models incorporating predicted secondary structure and accessible surface area and model performance evaluated by accuracy and MCC and validated on 28 independent prokaryotic glycoproteins.