GPCRpred: Prediction of families and superfamilies of GPCR

Welcome to the official documentation for GPCRpred, a specialized computational tool developed for the prediction and classification of G-protein coupled receptors (GPCRs). GPCRs are a major class of eukaryotic cell surface receptors involved in signal transduction and are critical targets for drug discovery. Web Server: [https://webs.iiitd.edu.in/raghava/gpcrpred] Citation Bhasin M, Raghava GPS (2004). GPCRpred: an SVM-based method for prediction of families and subfamilies of G-protein coupled receptors. Nucleic Acids Research, 32 (Web Server issue), W383-W389. https://doi.org/10.1093/nar/gkh416 GitHub:- (https://github.com/Manish-IIITD-repository/GPCRpred) About the Platform GPCRpred is a support vector machine (SVM) based method designed to classify GPCRs into their respective families and subfamilies. Given that GPCRs often share low sequence similarity, traditional alignment-based methods like BLAST may fail to provide accurate classifications. GPCRpred addresses this by utilizing global sequence descriptors. The platform is designed to: Functional Annotation: Assign uncharacterized protein sequences to specific GPCR families. Drug Target Identification: Facilitate the identification of receptors belonging to families of high therapeutic interest. Large-Scale Analysis: Process entire proteomes to find and classify potential GPCRs. Key Features Predictive Modeling Machine Learning: Utilizes Support Vector Machines (SVM) with a Radial Basis Function (RBF) kernel. Classification Hierarchy: Level 1: Distinguishes GPCRs from non-GPCRs. Level 2: Classifies GPCRs into five major families (A, B, C, D, and others). Level 3: Further classifies Family A into various subfamilies (e.g., Amine, Peptide, Hormone protein). High Performance: Achieved an overall accuracy of 97.3% for classifying the five major families of GPCRs. Achieved nearly 100% accuracy in distinguishing GPCRs from non-GPCRs using dipeptide composition. Feature Integration Dipeptide Composition: Encapsulates 400-dimensional vectors to capture the frequency and local order of amino acids, which is more informative than simple amino acid composition for GPCR classification. Robust Validation: Tested using 5-fold cross-validation on non-redundant datasets derived from GPCRDB. Overview of Model Development The training datasets were obtained from the GPCRDB (March 2003 release) and filtered for non-redundancy. The classification performance was measured using accuracy, sensitivity, specificity, and Matthews Correlation Coefficient (MCC). Classification Level Accuracy MCC GPCR vs. non-GPCR 99.5% 0.99 Major Families (A, B, C, D, etc.) 97.3% 0.90+ Family A Subfamilies 86.8% 0.85 Applications Pharmacology: Identifying new members of GPCR families for targeted drug screening. Proteomics: Annotating G-protein signaling components in eukaryotic genomes. Structural Biology: Predicting the functional class of receptors where experimental structures are unavailable. Contact & Support Prof. G.P.S. Raghava Head, Department of Computational Biology Indraprastha Institute of Information Technology (IIIT-Delhi), India. Email: [e-mail] License This research and associated software are distributed under the Creative Commons Attribution License, allowing for use and distribution with proper credit to the original authors.