Accelerating molecular docking calculations using graphics processing units.

Area of Science:

• Computational chemistry
• Molecular modeling
• Bioinformatics

Background:

• Molecular modeling is crucial for drug discovery and understanding biological systems.
• Protein-ligand and protein-protein docking require intensive computation for conformation generation and potential evaluation.
• Current CPU-based methods can be a bottleneck for large-scale molecular modeling tasks.

Purpose of the Study:

• To develop and evaluate a GPU-accelerated approach for molecular modeling tasks.
• To significantly enhance the speed of conformation generation and interaction potential evaluation.
• To integrate GPU acceleration into existing docking algorithms like PLANTS.

Main Methods:

• Implementation of a GPU-accelerated algorithm for molecular conformation generation and interaction potential evaluation.
• Optimization of the approach for rigid protein-protein docking and flexible protein-ligand docking scenarios.
• Parallelization of the PLANTS docking algorithm utilizing the GPU-accelerated scoring function.

Main Results:

• Achieved speedup factors of up to 50 for rigid protein-protein docking interaction potential evaluation.
• Observed speedup factors of up to 16 for flexible protein-ligand interaction potential evaluation.
• Demonstrated speedup factors of up to 10 and 7 for the PLANTS algorithm with varying ligand sizes and rotatable bonds, respectively.
• GPU-accelerated fitness landscape analysis in rigid protein-protein docking showed speedups up to 60.

Conclusions:

• GPU acceleration offers significant computational advantages for molecular modeling tasks.
• The developed approach substantially reduces computation time for protein-ligand and protein-protein docking.
• This GPU-accelerated method has the potential to accelerate drug discovery and biological structure analysis.