MatrixEpistasis

MatrixEpistasis performs exhaustive, covariate-adjusted genome-wide scans for SNP–SNP epistasis to identify interactions affecting quantitative traits.

Key Features:

• Exhaustive genome-wide scan: Performs an exhaustive search of SNP–SNP interactions across the genome.
• Covariate adjustment: Implements full covariate adjustment to mitigate confounding and improve power for detecting epistasis.
• Matrix-algebra formulation: Expresses exhaustive epistasis scanning and covariate adjustment as large matrix operations.
• Ultrafast computation: Achieves approximately 10,000× speed improvement over existing methods via matrix operations.
• Quantitative-trait focus: Targets detection of epistatic interactions contributing to quantitative trait variation.
• Validation: Validated using simulation data and real biological datasets.
• R implementation: Implemented in R.
• Empirical result: Re-analysis of a yeast dataset (11,623 SNPs, 1,008 segregants, 46 quantitative traits) identified thousands of putative epistatic interactions with P-values below 1.48e-10.

Scientific Applications:

• Epistasis discovery: Detection of SNP–SNP interactions that contribute to trait heritability and explain components of missing heritability.
• Covariate-aware genetic analysis: Identification of interactions while accounting for confounding covariates to improve inference accuracy.
• Re-analysis of genetic datasets: Large-scale re-analysis of organismal datasets (for example, yeast) to uncover novel putative interactions.

Methodology:

Derives mathematical formulas to express exhaustive epistasis scanning and full covariate adjustment as large matrix operations, validated on simulation and real datasets, and implemented in R.