A benchmark study of loop modeling methods applied to G protein-coupled receptors

Area of Science:

• Computational chemistry
• Structural biology
• Pharmacology

Background:

• G protein-coupled receptors (GPCRs) are critical drug targets, but many structures remain unsolved.
• Accurate prediction of extracellular loop 2 (ECL2) structures is challenging yet vital for ligand binding in Class A GPCRs.

Purpose of the Study:

• To evaluate the performance of different loop modeling protocols in Rosetta and MOE for GPCR ECL2 structure prediction.
• To identify optimal methods for improving ECL2 modeling accuracy in comparative modeling workflows.

Main Methods:

• Comparative modeling of GPCR ECL2 structures using Rosetta (CCD, KICF, NGK) and MOE (de novo search).
• Analysis of loop modeling performance based on loop length, with a focus on sub-angstrom and near-atomic accuracy.
• Utilizing conserved disulfide bonds (Cys3.25-Cys45.50) as a filtering strategy to enhance structure quality.

Main Results:

• Rosetta's KICF and NGK methods outperformed CCD and MOE's de novo search for ECL2 loops ≤24 residues, achieving higher accuracy.
• No tested method achieved near-atomic accuracy for longer ECL2 targets (25-32 residues), indicating a need for enhanced sampling.
• Implementing a disulfide bond distance filter (≤5.1 Å) significantly improved the quality of the top-ranked predicted models.

Conclusions:

• KICF and NGK are recommended for modeling shorter GPCR ECL2 segments.
• Advanced sampling techniques are required for accurate prediction of longer ECL2 loops.
• Disulfide bond constraints effectively refine GPCR ECL2 models, aiding virtual screening and drug design.