pulver: an R package for parallel ultra-rapid p-value computation for linear regression interaction terms.

Area of Science:

• Genetics
• Metabolomics
• Systems Biology

Background:

• Genome-wide association studies (GWAS) investigate complex disease genetics.
• Investigating associations between genetic variants and metabolite levels is crucial for understanding disease mechanisms.
• Existing tools lack methods for large-scale analysis of interactions between diverse 'omics' data types.

Purpose of the Study:

• To develop a practical tool for large-scale screening of interaction terms between arbitrary quantitative variables.
• To enable the investigation of interplay among multiple 'omics' layers, such as DNA methylation, genetic variants, and metabolite levels.

Main Methods:

• Developed the R package 'pulver' for computing p-values in numerous linear regression models.
• Implemented an algorithm in R and C++ that utilizes the correlation coefficient for hypothesis testing, enhancing speed.
• Optimized computation by avoiding costly inversions and rearranging iteration order through 'omics' layers.

Main Results:

• The 'pulver' package significantly outperforms existing tools in speed for large-scale interaction analysis.
• Demonstrated the package's utility by applying it to real-world data from the German KORA study.
• Successfully investigated the interplay among DNA methylation, genetic variants, and metabolite levels.

Conclusions:

• 'pulver' is a fast and convenient R package for screening a vast number of linear regression models for significant interaction terms.
• The tool facilitates the analysis of interactions between any pair of quantitative variables across different 'omics' datasets.
• 'pulver' is freely available on CRAN for researchers studying complex biological systems.