High-Efficiency Microiterative Optimization in QM/MM Simulations of Large Flexible Systems.

Abstract

We present here a double-optimizations-of-buffer-region (DOBR) microiterative scheme for high-efficiency energy minimizations of large, flexible systems in combined quantum-mechanical/molecular-mechanical (QM/MM) calculations. In the DOBR scheme, an entire system is divided into three regions: the core, buffer, and outer regions. The core region includes QM atoms and the MM atoms within a cutoff distance R1 to the QM atoms (denoted by MM1 atoms), and the buffer region consists of MM atoms within another cutoff distance R2 to MM1 atoms. Each DOBR microcycle involves two steps: First, QM atoms are assigned electrostatic-potential (ESP) charges, and the buffer and outer regions are optimized at the MM level with the core region kept frozen. Second, the core and buffer regions are optimized at the QM/MM level using the electrostatic embedding with the outer region kept frozen. The two steps are repeated until two optimizations converge at one structure. The DOBR scheme was tested in the optimizations of nucleobases solvated in water spheres of 30 Å radius, where the initial geometries were extracted from the trajectories of classical molecular dynamics simulations, and the cutoff distances R1 and R2 were set to 5.0 and 4.0 Å, respectively. For comparisons, the optimizations were also carried out by a "standard" scheme without microiteration and by the two-region microiterative (TRM) method. We found that the averaged number of QM calculations for the DOBR scheme is only ∼1% of that of the standard scheme and ∼6% of the TRM approach. The promising results indicate that the DOBR scheme could significantly increase the efficiency of geometry optimizations for large, flexible systems in QM/MM calculations.