Fine mapping and accurate prediction of complex traits using Bayesian Variable Selection models applied to...

Area of Science:

• Genetics
• Statistical genetics
• Bioinformatics

Background:

• Genome-Wide Association Studies (GWAS) are crucial for identifying genetic variants associated with diseases.
• Large sample sizes and high-density SNP data in modern GWAS reduce the mapping resolution of standard marginal association tests.
• There is a need for methods that can precisely map risk variants and predict polygenic risk scores in large-scale genetic studies.

Purpose of the Study:

• To introduce and evaluate Multi-locus Bayesian Variable Selection (BVS) as a powerful and precise method for GWAS.
• To compare the performance of BVS against marginal association tests in terms of mapping resolution, power, and false discovery rate.
• To demonstrate the utility of BVS for polygenic risk score (PRS) prediction using real-world biobank data.

Main Methods:

• Extensive simulations were conducted to assess the performance of multi-locus BVS methods.
• Bayesian Variable Selection (BVS) was applied to identify multiple genetic loci simultaneously.
• The methods were validated using blood biomarker data from the UK-Biobank, including approximately 300,000 samples and 5.5 million SNPs.
• Open-source R-software implementing the BVS methods was developed to handle biobank-scale data.

Main Results:

• Multi-locus BVS methods demonstrated high statistical power and a low false discovery rate in simulations.
• BVS achieved significantly better mapping resolution compared to traditional marginal association tests in GWAS.
• The study successfully applied BVS for both risk variant mapping and PRS prediction in a large UK-Biobank cohort.
• The accompanying R-software is scalable to biobank-sized datasets.

Conclusions:

• Multi-locus BVS is a superior method for fine-mapping genetic risk variants in large-scale GWAS.
• BVS provides a robust framework for accurate polygenic risk score prediction.
• The developed open-source software enables the application of advanced BVS methods to massive genetic datasets, facilitating genetic discovery.