Discovering epistasis in large scale genetic association studies by exploiting graphics cards

Area of Science:

• Genetics
• Computational Biology
• Bioinformatics

Background:

• Genome-wide association studies (GWAS) have generated vast amounts of germline variation data, yet explaining common disease heritability remains challenging.
• Current GWAS methods, focusing on single nucleotide polymorphisms (SNPs) independently, may not capture complex biological disease risk processes.
• Modeling genetic risk through SNP ensembles and interactions is a more biologically plausible approach but computationally intractable with standard methods.

Purpose of the Study:

• To analyze computational programs designed to overcome the challenges of genome-wide epistasis analysis using Graphics Processing Units (GPUs).
• To demonstrate the advantages of GPU technology for accelerating complex genetic interaction analyses.

Main Methods:

• In-depth analysis of software developed for GPU-accelerated epistasis screening.
• Tutorials on GPU technology explaining their computational advantages, particularly their high core density.
• Comparison of GPU-based methods against traditional CPU implementations for epistasis analysis.

Main Results:

• GPUs offer significant computational advantages due to their high density of microprocessor cores (e.g., over 2600 cores per GPU).
• Epistasis screens, which involve testing billions of models, are well-suited for GPU parallel processing.
• GPU implementations routinely achieve speedups exceeding two orders of magnitude (>100x) compared to standard CPU methods.

Conclusions:

• GPU acceleration is a powerful solution for the computational barriers in genome-wide epistasis analysis.
• The parallel processing capabilities of GPUs dramatically enhance the efficiency of identifying complex genetic interactions relevant to disease risk.
• Adoption of GPU technology can significantly advance gene mapping and our understanding of common disease heritability.