GlycoEP

GlycoEP predicts N-linked, O-linked, and C-linked glycosylation sites in eukaryotic glycoproteins to support accurate characterization of glycosites.

Key Features:

• Prediction targets: Identifies N-linked, O-linked, and C-linked glycosylation sites in eukaryotic glycoproteins.
• Datasets — standard: Uses a standard dataset curated so that no two glycosylated proteins share more than 40% sequence similarity.
• Datasets — advanced: Uses an advanced highly non-redundant dataset with a maximum of 60% similarity between any two glycosites' patterns.
• Machine-learning evaluation: Evaluated several algorithms using various machine-learning techniques to determine the most effective prediction approach.
• Algorithm: Employs Support Vector Machine (SVM) models selected as optimal from the evaluated methods.
• Performance metrics: Reported accuracies are 84.26% for N-linked, 86.87% for O-linked, and 91.43% for C-linked glycosites, with Matthews Correlation Coefficients of 0.54, 0.20, and 0.78 respectively.
• Sequon analysis: Provides analysis and prediction of sequons within input protein sequences.

Scientific Applications:

• Protein folding studies: Mapping glycosylation sites to investigate effects on protein folding and stability.
• Cell–cell interactions: Identifying glycosites that modulate cell–cell adhesion and signaling.
• Cell recognition: Characterizing glycans involved in cell recognition processes.
• Host–pathogen dynamics: Detecting glycosylation sites relevant to host–pathogen interactions.
• Therapeutic development and biotechnology: Informing design and characterization of glycoprotein therapeutics and biotechnological applications.
• Glycoprotein characterization: Comprehensive identification of glycosites for functional and structural studies.

Methodology:

Models were trained and evaluated using two curated datasets (standard: ≤40% protein similarity; advanced: ≤60% glycosite-pattern similarity), multiple machine-learning algorithms were tested, and Support Vector Machine (SVM) models were selected with reported accuracy and MCC values for N-, O-, and C-linked glycosites.