Accelerating electrostatic surface potential calculation with multi-scale approximation on graphics processing units.

Area of Science:

• Computational Biology
• Biophysics
• Scientific Computing

Background:

• Accurate computation of biomolecular electrostatic surface potential is crucial for understanding molecular function.
• Current methods for large biomolecules are computationally intensive, often requiring days on standard hardware.
• Existing acceleration techniques include multi-scale coarse-graining and parallel processing.

Purpose of the Study:

• To investigate the combined effect of multi-scale approximation and GPU parallelization on electrostatic potential computation speed.
• To demonstrate a synergistic speed-up achievable by integrating hierarchical charge partitioning (HCP) with graphical processing unit (GPU) acceleration.

Main Methods:

• Employed the analytical linearized Poisson-Boltzmann (ALPB) method for electrostatic potential calculation.
• Utilized the hierarchical charge partitioning (HCP) method for multi-scale approximation.
• Implemented parallelization on an ATI Radeon 4870 graphical processing unit (GPU).

Main Results:

• Achieved a 934-fold speed-up for a large viral capsid (476,040 atoms) using the combined HCP and GPU approach.
• The combined method yielded significantly greater speed-up than either HCP approximation (42-fold) or GPU parallelization (182-fold) alone.
• Demonstrated synergistic acceleration, where the combined techniques outperform individual optimizations.

Conclusions:

• The integration of multi-scale coarse-graining and GPU parallelization offers a powerful strategy for accelerating biomolecular electrostatic computations.
• This combined approach dramatically reduces the time required for analyzing large biomolecular systems, facilitating advanced research.
• The findings highlight the potential for synergistic performance gains in scientific computing through hybrid methodologies.