QUICK and Robust ESP and RESP Charges for Computational Biochemistry: Open-Source GPU Implementation.

Area of Science:

• Computational Chemistry
• Molecular Modeling
• High-Performance Computing

Background:

• Accurate partial charges are crucial for molecular simulations and force field parametrization.
• Traditional electrostatic potential (ESP) charge calculations can be sensitive to molecular orientation and grid density.
• Existing methods often lack robustness and efficiency for large-scale applications.

Purpose of the Study:

• To implement highly efficient ab initio electrostatic potential (ESP) calculations on graphics processing units (GPUs).
• To introduce a novel partial charge scheme robust against molecular orientation.
• To enable ultradense-grid ESP computations for improved accuracy and reliability.

Main Methods:

• Implementation of ESP calculations on GPUs using the Quantum Interaction Computational Kernel (QUICK) code.
• Development of a reweighted RESP (rwRESP) charge scheme to overcome sensitivity to grid point number.
• Performance analysis comparing GPU and CPU computational times.
• Validation of charge robustness against molecular orientation.

Main Results:

• A single data center GPU outperforms 128 CPU cores in ESP calculation time.
• Ultradense-grid ESP computations (∼20000 points/atom) achieve orientation independence.
• The novel rwRESP charge scheme demonstrates robustness against molecular orientation and grid density.
• Seamless integration of QUICK with AmberTools facilitates parametrization of GAFF for nonstandard residues.

Conclusions:

• The GPU-accelerated protocol provides highly efficient and robust molecular charge calculations.
• This method facilitates high-throughput parametrization and GPU-accelerated polarizable QM/MM simulations.
• The developed protocol serves as a foundational step for advanced molecular modeling applications.