iGPCR-Drug

iGPCR-Drug predicts interactions between G-protein-coupled receptors (GPCRs) and drug compounds to computationally identify GPCR–drug associations without requiring three-dimensional receptor structures.

Key Features:

• Sequence-based classifier: Uses sequence-derived features to predict GPCR–drug interactions rather than relying on 3D structural data.
• Drug representation: Represents drug compounds as two-dimensional fingerprints encoded as 256-dimensional vectors.
• GPCR representation: Characterizes GPCRs with pseudo amino acid composition (PseAAC) generated through grey model theory.
• Prediction algorithm: Employs a fuzzy K-nearest neighbour algorithm as the core prediction engine.
• Performance evaluation: Reports an overall success rate of 85.5% assessed by jackknife testing.
• Structure independence: Circumvents the need for experimentally determined GPCR three-dimensional structures.
• Extensibility: Methodology has potential to be applied to other drug–target interaction networks and high-throughput studies.

Scientific Applications:

• Therapeutic target identification: Prioritizes GPCR targets relevant to diseases such as cancer, diabetes, neurodegenerative disorders, inflammatory conditions, and respiratory ailments.
• Drug discovery support: Provides computational predictions to support basic research and drug development efforts focused on GPCRs.
• High-throughput interaction screening: Facilitates large-scale prediction of GPCR–drug associations for network-level studies.

Methodology:

Sequence-based classification using GPCR PseAAC derived via grey model theory and drug 2D fingerprints as 256-dimensional vectors, with interaction prediction by a fuzzy K-nearest neighbour algorithm and evaluation by jackknife testing (85.5% success rate), avoiding use of 3D structures.