Towards chemically accurate QM/MM simulations on GPUs.

Area of Science:

• Computational Chemistry
• Biophysics
• Molecular Modeling

Background:

• Computational chemistry simulations are vital for modeling natural phenomena.
• Hybrid QM/MM methods offer a balance between the performance of molecular mechanics and the accuracy of quantum mechanical calculations.
• Accurate molecular simulations are crucial in drug discovery and materials science.

Purpose of the Study:

• To evaluate a GPU-accelerated ONIOM implementation for hybrid QM/MM calculations.
• To investigate the trade-off between QM region size, energy accuracy, and simulation time.
• To demonstrate the convergence of ONIOM energy towards pure quantum mechanical energy.

Main Methods:

• Utilized a GPU-accelerated ONIOM implementation.
• Performed calculations on crambin and HIV integrase proteins.
• Employed the BrianQC quantum chemistry module for Hartree-Fock Self-Consistent Field (SCF) calculations.
• Used a GPU-accelerated MMFF94 force field for molecular mechanics computations.

Main Results:

• Larger QM regions in QM/MM calculations yield improved energy accuracy at the cost of increased simulation time.
• The ONIOM energy was shown to monotonically approach the pure quantum mechanical energy of the entire system.
• Demonstrated the feasibility of GPU acceleration for hybrid QM/MM simulations.

Conclusions:

• The choice of QM region size in QM/MM calculations presents a critical trade-off between accuracy and computational efficiency.
• GPU acceleration significantly enhances the performance of QM/MM simulations.
• The evaluated ONIOM implementation provides a robust approach for accurate and efficient molecular modeling.