NetOGlyc

NetOGlyc predicts mucin-type O-glycosylation sites in proteins by using artificial neural networks to analyze sequence context and surface accessibility for the identification of site-specific O-linked glycosylation.

Key Features:

• Neural network ensemble: Employs a jury of artificial neural networks to evaluate potential O-glycosylation sites based on sequence context.
• Prediction performance: Reports correct identification of 83% of glycosylated residues and 90% of non-glycosylated residues on independent test sets.
• Acceptor-site discrimination: Distinguishes between serine and threonine acceptor sites to account for differing sequence contexts.
• Sequence pattern recognition: Identifies patterns such as clustering of glycosylation sites and preferences for particular amino acids at positions relative to the modified residue.
• Surface accessibility: Incorporates surface accessibility information alongside sequence context in predictions.
• Curated training data and database links: Is informed by data from O-GLYCBASE with cross-references to SWISS-PROT and PDB.

Scientific Applications:

• Proteome-wide glycosylation mapping: Predicts O-glycosylation sites across proteomes to support studies of site-specific regulation of protein function.
• Viral protein analysis: Predicts O-glycosylation sites on viral envelope proteins such as gp120 from HIV-1, HIV-2, and SIV to inform analyses of conserved glycosylation signals.
• Glycoprotein functionality studies: Maps glycosylation sites to aid investigation of effects on protein stability, folding, and molecular interactions.

Methodology:

Uses a jury of neural networks trained on a dataset of 48 mammalian glycoproteins containing 264 O-glycosylation sites, analyzes sequence context extending beyond the −4 to +4 region, and incorporates surface accessibility while recognizing features such as proline abundance and clustering of sites.