GlycoPP: Prediction of potential N-and O-glycosites in prokaryotic proteins

Welcome to the official repository and documentation overview for GlycoPP, an open-access platform designed for the accurate prediction of N- and O-linked glycosites in prokaryotic protein sequences. Developed to address the unique structural and sequence contexts of Archaea and Bacteria, GlycoPP provides a specialized tool for identifying these critical post-translational modifications (PTMs). Web Server: http://webs.iiitd.edu.in/raghava/glycopp/ Citation Chauhan JS, Bhat AH, Raghava GPS, Rao A (2012). GlycoPP: A Webserver for Prediction of N- and O-Glycosites in Prokaryotic Protein Sequences. PLoS ONE 7(7): e40155. https://doi.org/10.1371/journal.pone.0040155 GitHub:-https://github.com/Manish-IIITD-repository/GlycoPP About the Platform GlycoPP is the first web server specifically trained on prokaryotic glycoproteins, recognizing that eukaryotic prediction tools often fail when applied to bacterial and archaeal sequences[cite: 3]. The platform utilizes Support Vector Machine (SVM) algorithms to distinguish between glycosylated and non-glycosylated residues based on extensive experimental data. The platform supports: N-linked Glycosylation: Glycan attachment to the amide group of Asparagine (Asn). O-linked Glycosylation: Glycan attachment to the hydroxyl group of Serine (Ser), Threonine (Thr), or Tyrosine (Tyr. Key Features Predictive Modeling Machine Learning: Built using SVMlight with various kernels, optimized to handle the specific sequence motifs of prokaryotes. Feature Diversity: Incorporates Binary Profiles (BPP), Composition Profiles (CPP), and Position-Specific Scoring Matrices (PSSM/PPP). Accuracy: Achieved an accuracy of 82.71% (MCC 0.65) for N-glycosites and 73.71% (MCC 0.48) for O-glycosites. Structural & Hybrid Models Post-Translational Insight: Since prokaryotic glycosylation often occurs on folded proteins, models include predicted secondary structure and surface accessibility. Surface Accessibility: GlycoPP uses ASA (Accessible Surface Area) values to improve prediction, as glycosites are predominantly found on flexible, exposed loops. Data Resources Extensive Training Set: Derived from 59 experimentally characterized glycoproteins, including 107 N-linked and 116 O-linked glycosites. Wide Domain Coverage: Includes data from phyla such as Proteobacteria, Crenarchaeota, Firmicutes, and Bacteroidetes. Overview of Model Development GlycoPP models were validated using 5-fold cross-validation and tested against an independent dataset of 28 prokaryotic glycoproteins to ensure reliability across different phyla. Feature Model N-linked Accuracy O-linked Accuracy BPP (Binary Profile) 79.91% 66.81% PPP (PSSM Profile) 73.11% 73.28% Hybrid (BPP + ASA) 82.71% 69.40% Hybrid (PPP + SS + ASA) 73.83% 73.71% Applications Pathogen Research: Understanding host-pathogen interactions mediated by bacterial glycoproteins. Vaccine Development: Identifying immunogenic glycoproteins in pathogens like Campylobacter jejuni or Mycobacterium tuberculosis. Evolutionary Biology: Gaining insights into the acceptor specificities of prokaryotic glycosyltransferases. Contact & Authors Prof. G.P.S. [e-mail] Centre, Indraprastha Institute of Information Technology (IIIT-Delhi), India. License This platform and its associated research are distributed under the Creative Commons Attribution License, allowing for unrestricted use and distribution with proper credit to the original authors.