A Computational Framework for GPCR Ligand Discovery
Year: 2026
Type: software
Abstract: Two complementary computational pipelines for prioritizing analogs of known GPCR ligands, and code for comparing their outputs. The first pipeline combines P2Rank binding-site detection with independent AutoDock Vina docking into three predicted pockets, ranking compounds by a cross-pocket consensus vote. The second combines homology-guided orthosteric docking with Random Forest and multilayer perceptron QSAR models trained on ChEMBL bioactivity data across 16 Class A GPCR targets, ranking compounds by a combined docking and QSAR score. A third notebook compares the two rankings by exact canonical SMILES and by shared ring chemotype. Neither pipeline reads the other's results; the comparison is performed afterwards. The framework is demonstrated on an orphan Class A GPCR of Caenorhabditis elegans, starting from two of its known ligands. This repository accompanies the manuscript "A Computational Framework for GPCR Ligand Discovery" (Cabrera, Strickland, Vela and Jia).
Authors: Cabrera, Itzel, Strickland, David, Vela, Michelle, Jia, Kailiang
Institution: Florida Atlantic University
Language: English
Subfield: Computational Theory and Mathematics
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7213766190