Related Experiment Videos
Bayesian modelling of multivariate quantitative traits using seemingly unrelated regressions
Claudio J Verzilli1, Nigel Stallard, John C Whittaker
1Department of Epidemiology and Public Health, Imperial College London, London, United Kingdom. c.verzilli@imperial.ac.uk
Genetic Epidemiology
|March 25, 2005
Summary
This study introduces a Bayesian approach using Seemingly Unrelated Regressions (SUR) for analyzing genetic markers and multiple traits. The Bayesian SUR method improves model selection accuracy for complex genetic associations compared to traditional univariate analyses.
Area of Science:
- Statistical Genetics
- Bioinformatics
- Computational Biology
Background:
- Understanding the statistical associations between genetic markers and complex traits is crucial in genomics.
- Multivariate quantitative traits often exhibit complex interdependencies influenced by multiple genetic loci.
- Existing methods may not fully capture the intricate relationships between multiple genetic markers and several phenotypes simultaneously.
Purpose of the Study:
- To develop and evaluate a Bayesian approach for modeling the statistical association between multi-locus genetic markers and multivariate quantitative traits.
- To introduce and demonstrate the utility of Bayesian Seemingly Unrelated Regressions (SUR) for this purpose.
- To compare the performance of the Bayesian SUR approach against univariate analyses in terms of model selection accuracy.
Main Methods:
- Utilized Bayesian Seemingly Unrelated Regressions (SUR) to model associations between genotypes at different loci and multiple phenotypes.
- Allowed for non-simultaneous effects of genotypes on phenotypes and assumed correlated residuals between regressions.
- Employed simulations to assess the probability of selecting the true model under general conditions.
- Applied the method to real data, including imputation of missing genotype data using a nested haplotype phasing algorithm.
Main Results:
- Simulation results indicate that the Bayesian SUR approach offers an increased probability of selecting the true model compared to univariate analyses.
- The developed Bayesian framework naturally accommodates missing genotype data through imputation.
- The method was successfully applied to analyze 12 SNPs in the apolipoprotein E (APOE) gene and their association with lipid level changes in response to atorvastatin treatment.
Conclusions:
- The Bayesian SUR approach provides a robust framework for modeling complex genetic associations with multivariate quantitative traits.
- This method enhances model selection accuracy, particularly in scenarios with correlated phenotypes and multi-locus effects.
- The approach effectively handles missing data, making it suitable for real-world genetic studies.