BGWAS: Bayesian variable selection in linear mixed models with nonlocal priors for genome-wide association studies

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genome-wide association studies (GWAS) identify single nucleotide polymorphisms (SNPs) linked to phenotypes.
• High false positive rates are common in GWAS due to correlated data and large SNP numbers.

Purpose of the Study:

• To introduce BGWAS, a Bayesian variable selection method for GWAS.
• To address high false positive rates in GWAS using nonlocal priors in linear mixed models (LMMs).

Main Methods:

• BGWAS employs a two-step approach: SNP screening and LMM-based model selection.
• The screening step uses Bayesian false discovery control for candidate SNP selection.
• Model selection explores LMMs with varying numbers of candidate SNPs.

Main Results:

• Simulations show BGWAS significantly reduces false positives compared to existing GWAS methods.
• BGWAS maintains or improves the detection of true positive SNPs.
• Case studies demonstrate BGWAS utility in plant salt stress and alcohol use disorder.

Conclusions:

• BGWAS effectively lowers false positives in GWAS.
• The method preserves and can enhance the identification of true positive SNPs.
• BGWAS offers a flexible and accurate approach for genetic association studies.