Simultaneous Bayesian analysis of contingency tables in genetic association studies

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genetic association studies analyze genotype-phenotype relationships using contingency tables.
• Case-control designs involve fixed row counts and random column counts in these tables.
• The primary goal is to test for independence between phenotype and genotype at each genetic locus.

Purpose of the Study:

• To present an objective Bayesian methodology for genetic association tests.
• To provide an alternative to frequentist methods by utilizing Dirichlet-multinomial conjugacy.
• To address the challenge of specifying prior probabilities in multiple testing scenarios.

Main Methods:

• Developed an objective Bayesian methodology based on Dirichlet-multinomial conjugacy.
• Applied the Bayesian approach to The Wellcome Trust Case Control Consortium (WTCCC) data.
• Incorporated linkage disequilibrium and effective number of tests for prior probability specification.
• Utilized a Bayesian decision theoretic multiple test procedure.

Main Results:

• The Bayesian tests, based on the likelihood principle, avoid computationally intensive methods.
• Bayes factor ordering demonstrated good agreement with frequentist p-values.
• The methodology effectively handles prior probability specification in complex genetic data.
• Demonstrated successful application to large-scale WTCCC data.

Conclusions:

• The proposed Bayesian methodology offers a valid and efficient approach for genetic association testing.
• The method provides a robust framework for analyzing genotype-phenotype associations, especially in large datasets.
• Reconciliation strategies for frequentist and Bayesian approaches in multiple association testing were discussed.