Seminário no âmbito do Mestrado em Bioestatística, por Miguel Pereira (National Heart and Lung Institute, Imperial College London, UK).

Genome-Wide Association Studies are usually analysed by estimating SNP effects individually (standard analysis) and adjusting for multiple testing. However, SNPs detected by this method explain only a small fraction of the predicted heritability for most traits. Here we aim to improve SNP detection by integrating external biological information about the SNPs in a Bayesian hierarchical shrinkage model that jointly estimates SNP effects. SNP effects are assumed to follow a normal distribution centered at zero and prior biological information retrieved from online databases is used to modulate the variance of the SNP effects through differential shrinkage. SNPs with more biological support are less shrunk towards zero, thus being more likely detected.
In this talk, I will present the approach of differential shrinkage to integrate external information in a Bayesian model and show its improvement comparing to the standard statistical approach. Also, I will present a novel measure to determine the best shrinkage parameters in this setting and how it can be applied to other models.