Incorporating Prior Knowledge about Genetic Variants into the Analysis of Genetic Association Data: An Empirical...

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Local false discovery rate (LFDR) in genome-wide association studies (GWAS) is sensitive to the choice of reference SNP class.
• LFDR values can differ significantly when calculated against separate (e.g., exonic) versus combined (all SNPs) reference classes.
• This variability can lead to conflicting conclusions about SNP-disease associations.

Purpose of the Study:

• To develop novel empirical Bayes methods for LFDR estimation in GWAS.
• To address the ambiguity arising from different reference class choices.
• To improve the reliability and accuracy of SNP-disease association findings.

Main Methods:

• Introduction of empirical Bayes methods that integrate both separate and combined SNP reference classes.
• Development of a maximum entropy approach that adaptively utilizes reference classes based on data reliability.
• Application of these methods to GWAS data from 2,000 cases and 3,000 controls.

Main Results:

• Simulation studies demonstrated improved performance of the proposed empirical Bayes methods.
• The maximum entropy method showed effective adaptation, prioritizing separate classes when data is sufficient and combined classes otherwise.
• Analysis of real GWAS data using the new methods provided refined association insights.

Conclusions:

• The proposed empirical Bayes methods offer a more robust approach to LFDR estimation in GWAS.
• These methods enhance the accuracy of identifying disease-associated single nucleotide polymorphisms (SNPs).
• R functions and a Shiny app are available for implementing these advanced statistical techniques.