A rapid gene-based genome-wide association test with multivariate traits

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Gene-based genome-wide association studies (GWAS) offer advantages over single nucleotide polymorphism (SNP) analyses by reducing multiple testing burdens and increasing power.
• Analyzing multivariate traits in gene-based GWAS presents significant analytical and computational challenges.

Purpose of the Study:

• To introduce a rapid implementation of multivariate multiple linear regression (RMMLR) for gene-based GWAS.
• To enable efficient analysis of both single and multivariate traits at a genome-wide level.

Main Methods:

• The RMMLR approach was developed for unrelated individuals and families, incorporating covariates.
• The method's test statistic distribution is robust to linkage disequilibrium (LD) among SNPs.
• An R package was created to facilitate multivariate gene-based GWAS using the RMMLR approach.

Main Results:

• Simulations demonstrated that RMMLR maintains correct type I error rates, even with SNPs in strong LD.
• The RMMLR approach showed increased power in detecting genes associated with a subset of traits.
• Performance was evaluated using the Minnesota Center for Twin Family Research dataset.

Conclusions:

• The RMMLR approach is an efficient and powerful tool for gene-based GWAS.
• It accurately performs analyses for both single and multivariate traits while maintaining type I error control.