pdCSM-GPCR

pdCSM-GPCR predicts potent small-molecule ligands for G protein-coupled receptors (GPCRs) using graph-based signatures to support identification and prioritization of GPCR ligands in drug discovery.

Key Features:

• Graph-Based Signatures: Uses graph-based signatures to model ligand–GPCR interactions.
• Comprehensive Predictive Models: Built from curated bioactivity data (IC50, EC50, Ki, and Kd) for 36 GPCR targets across classes A, B, C, and F.
• High Performance: Validated by stratified 10-fold cross-validation and blind tests achieving Pearson correlations up to 0.89 and outperforming previous methods.
• Feature Identification: Identified structural features associated with potent GPCR ligands, including bicyclic rings and increased aromaticity.

Scientific Applications:

• Drug Discovery and Development: Screens and prioritizes compound libraries for potential GPCR ligands to guide experimental validation.
• Research and Analysis: Provides insights into structural determinants of ligand potency to support studies of GPCR–ligand interactions.

Methodology:

Curated bioactivity data (IC50, EC50, Ki, and Kd) for 36 GPCR targets across classes A, B, C, and F and used graph-based signatures to construct predictive models evaluated by stratified 10-fold cross-validation and blind tests.