Exploring predictive biomarkers from clinical genome-wide association studies via multidimensional hierarchical mixture models

摘要

Workflow of the gene screening method using multidimensional hierarchical mixture models. The top panels show the genotypic and phenotypic data of control and treatment groups. The sample sizes of control and treatment groups are denoted by and , and is the total number of SNPs. The middle panels show the summary statistics consisting of estimated effect sizes , and their standard errors for the main effect of SNPs derived from association analyses using regression models without interaction terms. The bottom panels show the denoised effect size distribution of SNPs that are associated with outcomes, posterior probabilities of association, and adjusted effect size estimates for each SNP. Significant SNPs are detected by the optimal discovery procedure based on the distribution and posterior probabilities