BICOSS: Bayesian iterative conditional stochastic search for GWAS.

Area of Science:

• Genetics
• Statistical genomics
• Bioinformatics

Background:

• Single marker analysis (SMA) is widely used in genome-wide association studies (GWAS) but suffers from poor false discovery control and low power for small/medium effect variants.
• This limitation hinders the identification of true causal single nucleotide polymorphisms (SNPs).

Purpose of the Study:

• To introduce a novel method, Bayesian Iterative Conditional Stochastic Search (BICOSS), designed to enhance GWAS.
• BICOSS aims to improve false discovery rate control and increase the recall of variants with small and medium effect sizes.

Main Methods:

• BICOSS employs an iterative approach, alternating between a screening step and a Bayesian model selection step.
• This methodology is designed for robust statistical inference in genetic association studies.

Main Results:

• Simulation studies demonstrate that BICOSS significantly reduces the false discovery rate compared to SMA.
• BICOSS enhances the statistical power to detect smaller effect sizes, leading to higher recall of true causal SNPs.
• Real-world applications highlight BICOSS's practical utility and adaptability.

Conclusions:

• BICOSS offers superior performance over traditional SMA in GWAS.
• The method achieves higher recall of true SNPs while substantially reducing the false discovery rate.