CARAT-GxG: CUDA-Accelerated Regression Analysis Toolkit for Large-Scale Gene-Gene Interaction with GPU Computing...

Area of Science:

• Genetics
• Computational Biology
• Bioinformatics

Background:

• Regression analysis is standard for identifying associations between phenotypes and genetic variants in genome-wide association studies (GWAS).
• Computational demands limit regression analysis to single markers, hindering the study of multiple single nucleotide polymorphisms (SNPs) and gene-gene interactions (GGI) in large datasets.
• Existing methods struggle with the computational complexity of analyzing GGI within large-scale GWAS data.

Purpose of the Study:

• To introduce CARAT-GxG, a novel toolkit designed for efficient regression analysis of GGI in GWAS.
• To leverage GPU computing and CUDA for significant performance enhancement in genetic association studies.
• To overcome the computational limitations of analyzing multiple SNPs and their interactions.

Main Methods:

• Development of CARAT-GxG, a toolkit optimized for GPU computing systems using CUDA.
• Implementation of GPU-specific optimization techniques to accelerate regression analysis.
• Integration with the TORQUE Resource Manager for scalable GPU computing.

Main Results:

• CARAT-GxG demonstrated a substantial execution speed increase, achieving nearly 700-fold acceleration compared to existing methods.
• The toolkit delivered highly reliable results, validating its accuracy for genetic association analysis.
• Near-linear speed acceleration was achieved through the GPU computing system implementation.

Conclusions:

• CARAT-GxG significantly enhances the feasibility of large-scale regression analysis for GGI in GWAS.
• The toolkit offers a powerful solution for computationally intensive genetic analyses.
• CARAT-GxG is expected to facilitate deeper insights into complex genetic architectures through the analysis of multiple SNPs and interactions.