The number of solved
G-protein-coupled receptor (GPCR) crystal structures has expanded
rapidly, but most GPCR structures remain unsolved. Therefore, computational
techniques, such as homology modeling, have been widely used to produce
the theoretical structures of various GPCRs for structure-based drug
design (SBDD). Due to the low sequence similarity shared by the transmembrane
domains of GPCRs, accurate prediction of GPCR structures by homology
modeling is quite challenging. In this study, angiotensin II type
I receptor (AT1R) was taken as a typical case to assess the reliability
of class A GPCR homology models for SBDD. Four homology models of
angiotensin II type I receptor (AT1R) at the inactive state were built
based on the crystal structures of CXCR4 chemokine receptor, CCR5
chemokine receptor, and δ-opioid receptor, and refined through
molecular dynamics (MD) simulations and induced-fit docking, to allow
for backbone and side-chain flexibility. Then, the quality of the
homology models was assessed relative to the crystal structures in
terms of two criteria commonly used in SBDD: prediction accuracy of
ligand-binding poses and screening power of docking-based virtual
screening. It was found that the crystal structures outperformed the
homology models prior to any refinement in both assessments. MD simulations
could generally improve the docking results for both the crystal structures
and homology models. Moreover, the optimized homology model refined
by MD simulations and induced-fit docking even shows a similar performance
of the docking assessment to the crystal structures. Our results indicate
that it is possible to establish a reliable class A GPCR homology
model for SBDD through the refinement by integrating multiple molecular
modeling techniques.