Ultra-fast FFT protein docking on graphics processors.

Area of Science:

• Structural bioinformatics
• Computational biology
• Drug discovery

Background:

• Protein-protein interactions (PPIs) are crucial in structural bioinformatics.
• In silico prediction of PPIs via docking is computationally intensive.
• Faster docking tools are needed for large-scale PPI network analysis and drug design.

Purpose of the Study:

• To implement the Hex protein docking algorithm on graphics processor units (GPUs).
• To evaluate the computational speed-up achieved by GPU acceleration for protein docking.
• To assess the suitability of the Hex algorithm for GPU exploitation.

Main Methods:

• Implementation of the Hex spherical polar Fourier protein docking algorithm on Nvidia GPUs.
• Utilizing multiple 1D fast Fourier transforms (FFTs) for 6D docking searches.
• Comparison of GPU performance against CPU-based calculations and other GPU-accelerated tools (ZDOCK, PIPER).

Main Results:

• An exhaustive 6D docking search was completed in 15 seconds on a GTX 285 GPU.
• This represents a 45-fold speed-up over single CPU calculations.
• The Hex algorithm demonstrates significant speed-up (orders of magnitude) compared to CPU and other GPU methods, particularly with 1D FFTs.

Conclusions:

• The Hex algorithm, when implemented on GPUs using 1D FFTs, enables exhaustive protein docking calculations in seconds.
• This GPU acceleration offers substantial computational advantages over traditional CPU-based methods.
• The Hex algorithm is highly suited for GPU acceleration, outperforming conventional 3D FFT docking approaches.