Genetic Inheritance and Genome Wide Association Statistical Test Performance | Abstract
Journal of Proteomics & Bioinformatics

Journal of Proteomics & Bioinformatics
Open Access

ISSN: 0974-276X


Genetic Inheritance and Genome Wide Association Statistical Test Performance

Philip Cooley, Robert Clark, Ralph Folsom and Grier Page

The choice of a statistical method significantly affects the power profiles of Genome Wide Association (GWA) predictions. Previous simulation studies of a single synthetic phenotype marker determined that the gene model or mode of inheritance (MOI) was a major influence on power. In this paper, the authors compare the power profiles of GWA statistical methods that combine MOI specific methods into multiple test scenarios against individual methods that may or not assume a MOI gene model consistent with the marker that predicts the association. Combining recessive, additive and dominant individual tests, and using either the Bonferroni Correction method or the MAX test (Li et al., 2008) has power implications with respect to single test GWA-based methods. If the gene model behind the associated phenotype is not known, a multiple test procedure could have significant advantages with respect to single test procedures. Our findings do not provide a specific answer as to which statistical method is best. The best method depends on the MOI gene model associated with the phenotype (diagnosis) in question. However, our results do indicate that the common assumption that the MOI of the locus associated with the diagnosis is additive has consequences. Our results indicate that researchers should consider a multi-test procedure that combines the results of individual MOI-based core tests as a statistical method for conducting the initial screen in a GW study. The process for combining the core tests into a single operational test can occur in a number of ways. We identify two: the Bonferroni procedures and the MAX procedure, each of which produce very similar statistical power profiles.