Nglyc

Nglyc predicts N-glycosylation sites in eukaryotic protein sequences using a Random Forest classifier to distinguish glycosylated N-X-[S/T] sequon motifs (X ≠ proline) from non-glycosylated sites.

Key Features:

• Random Forest classifier: Uses a Random Forest algorithm to classify N-glycosylation sites.
• Feature set (315 features): Leverages 315 features derived from the protein sequence and its homologs.
• Training dataset: Trained on a balanced dataset of 600 known glycosylated and 600 non-glycosylated sites.
• Target motif: Focuses on the N-X-[S/T] sequon motif where X is any amino acid except proline.
• Performance: Achieved training accuracy 0.8033 (using all features), testing accuracy 0.8248, sensitivity 0.8305, and specificity 0.8182.
• Benchmarking: Evaluated against NetNGlyc, EnsembleGly, and GPP and demonstrated superior accuracy, sensitivity, and specificity.
• Cross-species validation: Validated using human and mouse N-glycosylation sites.

Scientific Applications:

• N-glycosylation site prediction: Identification and annotation of potential N-glycosylation sites in eukaryotic proteins.
• Mechanistic studies: Investigating determinants that distinguish glycosylated from non-glycosylated N-X-[S/T] sequons.
• Method comparison and validation: Providing a benchmark for comparison with NetNGlyc, EnsembleGly, and GPP and for validating predictions across human and mouse datasets.

Methodology:

Random Forest classifier trained on 315 sequence- and homolog-derived features using a balanced dataset of 600 glycosylated and 600 non-glycosylated sites, with performance evaluated against NetNGlyc, EnsembleGly, and GPP and validated on human and mouse N-glycosylation sites.