EpiGPU: exhaustive pairwise epistasis scans parallelized on consumer level graphics cards

Area of Science:

• Genetics
• Computational Biology
• Bioinformatics

Background:

• Genome-wide association studies (GWAS) are common, but epistasis analysis is computationally prohibitive.
• Increased SNP chip density exacerbates computational burden for epistasis detection.
• Lack of empirical evidence for epistasis is due to analysis challenges.

Purpose of the Study:

• To develop a computationally efficient method for detecting epistasis.
• To make epistasis analysis accessible across various hardware platforms.
• To overcome the computational limitations hindering epistasis research.

Main Methods:

• Utilized graphics processing unit (GPU) multicore architecture to partition the epistasis search space.
• Implemented an exhaustive pairwise epistasis scan for quantitative phenotypes.
• Developed the solution using OpenCL for cross-platform and cross-vendor compatibility.

Main Results:

• Achieved a 92x speed increase for epistasis scans using a consumer GPU.
• Demonstrated that comparable speedup without a GPU would require a costly compute cluster.
• The OpenCL implementation is free, open-source, and platform-independent.

Conclusions:

• GPU acceleration significantly enhances the feasibility of epistasis analysis.
• This tool democratizes epistasis detection, making it viable for researchers without supercomputing resources.
• The developed software facilitates empirical investigation into the role of epistasis in complex traits.