Geometry Optimization\nUsing the Frozen Domain and\nPartial Dimer Approaches in the Fragment Molecular Orbital Method:\nImplementation, Benchmark, and Applications to Protein Ligand-Binding\nSites

The frozen domain (FD) approximation with the fragment\nmolecular\norbital (FMO) method is efficient for partial geometry optimization\nof large systems. We implemented the FD formulation (FD and frozen\ndomain dimer [FDD] methods) already proposed by Fedorov, D. G. et\nal. (<i>J. Phys. Chem. Lett.</i> <b>2011</b>, 2, 282–288);\nproposed a variation of it, namely frozen domain and partial dimer\n(FDPD) method; and applied it to several protein–ligand complexes.\nThe computational time for geometry optimization at the FDPD/HF/6-31G*\nlevel for the active site (six fragments) of the largest β<sub>2</sub>-adrenergic G-protein-coupled receptor (440 residues) was\nalmost half that of the conventional partial geometry optimization\nmethod. In the human estrogen receptor, the crystal structure was\nrefined by FDPD geometry optimization of estradiol, surrounding hydrogen-bonded\nresidues and a water molecule. The rather polarized ligand binding\nsite of influenza virus neuraminidase was also optimized by FDPD optimization,\nwhich relaxed steric repulsion around the ligand in the crystal structure\nand optimized hydrogen bonding. For Serine-Threonine Kinase Pim1 and\nsix inhibitors, the structures of the ligand binding site, Lys67,\nGlu121, Arg122, and benzofuranone ring and indole/azaindole ring of\nthe ligand, were optimized at FDPD/HF/6-31G* and the ligand binding\nenergy was estimated at the FMO-MP2/6-31G* level. As a result of examining\nthree different optimization regions, the correlation coefficient\nbetween pIC<sub>50</sub> and ligand binding energy was considerably\nimproved by expanding the optimized region; in other words, better\nstructure–activity relationships was obtained. Thus, this approach\nis promising as a high-precision structure refinement method for structure-based\ndrug discovery.