GlycoEP: In silico Platform for Prediction of Glycosites

Welcome to the official repository and documentation overview for GlycoEP, an open-access web server designed for the accurate prediction of N-, O-, and C-linked glycosites in eukaryotic protein sequences. Glycosylation is a critical post-translational modification involved in protein folding, cell-cell interactions, and host-pathogen recognition. GlycoEP was developed to provide biologists with a more robust tool for identifying these sites using high-quality, non-redundant datasets.

Web Server: [(https://webs.iiitd.edu.in/raghava/glycoep/suppli.html)]

Chauhan JS, Rao A, Raghava GPS (2013). In silico Platform for Prediction of N-, O- and C-Glycosites in Eukaryotic Protein Sequences. PLoS ONE 8(6): e67008. https://doi.org/10.1371/journal.pone.0067008

GlycoEP addresses the limitations of existing glycosylation prediction tools by utilizing larger, more stringently filtered datasets and advanced machine-learning techniques[cite: 1]. The platform supports the prediction of three major types of eukaryotic glycosylation:

• N-linked: Attachment to the Nitrogen atom of Asparagine (Asn) in the Asn−X−Ser/Thr sequon.
• O-linked: Attachment to Serine (Ser) or Threonine (Thr) residues.
• C-linked: Attachment to the first Tryptophan (Trp) in motifs like W−X−X−W.

The server offers two tiers of prediction models:

1. Standard Models: Developed from non-redundant proteins where no two sequences share more than 40% similarity.
2. Advanced Models: Developed from highly non-redundant pattern datasets where no two glycosite patterns share more than 60% similarity, ensuring more robust and biologically meaningful predictions.

• Machine Learning: Support Vector Machine (SVM) was identified as the optimum tool for these models, utilizing the RBF kernel.
• High Accuracy: Advanced SVM models achieved accuracies of 84.26% for N-linked, 86.87% for O-linked, and 91.43% for C-linked glycosites.
• Feature Integration: Models incorporate Binary Profiles of Patterns (BPP), Composition Profiles (CPP), and PSSM profiles (PPP).

• Secondary Structure: Incorporates three-state structure information (Coil, Helix, Strand) obtained via PSIPRED.
• Surface Accessibility: Includes Predicted Accessible Surface Area (ASA) values, which significantly improve O-glycosite prediction.

• Sequon Scanner: A dedicated tool to scan input protein sequences for both universal and rare N-linked glycosylation motifs.
• Multiple Submissions: The server allows users to submit multiple sequences simultaneously for high-throughput analysis.

GlycoEP models were trained and evaluated using a 5-fold cross-validation procedure. The datasets were derived from the SWISS-PROT database (June 2011 release), focusing exclusively on experimentally verified eukaryotic glycoproteins. Overlapping patterns of 21 residues were generated for each potential glycosite to capture the local sequence context.

• Proteomics: Characterization and annotation of glycosites in eukaryotic proteins.
• Therapeutics: Analysis of human therapeutic glycoproteins, which constitute over 70% of current protein-based drugs.
• Biological Research: Investigating the role of glycans in protein folding, cell signaling, and host-pathogen interactions.

Developed under the Open Source Drug Discovery (OSDD) initiative

Prof. Dr. Gajendra PS Raghava

raghava@iiitd.ac.in

This platform and its associated research are distributed under the Creative Commons Attribution License (CC BY 2.0).