SPRINT-Gly

SPRINT-Gly predicts N- and O-linked glycosylation sites in human and mouse proteins using machine-learning models to inform studies of protein structure–function and disease-associated glycosylation.

Key Features:

• Extensive dataset: Leverages a comprehensive dataset of N- and O-linked glycosylation sites from various sources curated for human and mouse proteins.
• Input features: Integrates sequence-based data, predicted structural information, and physicochemical properties of neighboring residues around candidate sites.
• Predictive models: Employs deep learning neural networks and support vector machines trained separately for human and mouse proteins.
• Performance: Reported an 18% higher Matthew's correlation coefficient for N-linked and a 50% increase for O-linked glycosylation predictions compared with existing methods.

Scientific Applications:

• Proteomics: Enables large-scale annotation of glycosylation sites to support protein characterization and structural analysis.
• Protein structure–function studies: Provides site-specific glycosylation predictions to help interpret effects on protein folding and function.
• Disease mechanism research: Supports investigation of altered glycosylation patterns implicated in human disease.
• Therapeutic target identification: Assists identification of glycosylation sites relevant to development of glycosylation-targeted interventions.

Methodology:

Models are trained on a curated dataset using sequence-based attributes, predicted structural features, and physicochemical properties of neighboring residues, with classifiers consisting of deep learning neural networks and support vector machines trained separately for human and mouse proteins.