Biostatistics Seminar Series: Yize Zhao, PhD

Title: Integrative Bayesian models for Imaging Genetics Abstract: Recent advances in genomic technologies accelerate the investigation of genetic underpinnings of complex phenotypes. The emergence of complex and large-scaled datasets further provide platforms to conduct data integration, such as linking genetic variants to multi-/high- dimensional brain imaging traits which provides valuable insights to both discover disease risk related genes and characterize their impacts on neural systems. In this talk, we present two relevant topics on the joint analysis of large-scale imaging genetics data motived by Alzheimers Disease Neuroimaging Initiative (ADNI). In the first part, based on genome-wide association studies (GWAS), we develop a computationally efficient Bayesian model based on multiresolution feature selection prior to identify risk imaging features and their associated genetic factors. Grouping information among SNPs and network configuration across brain regions are also incorporated in the model. In the second one, we target on genome-wise complex trait analysis (GCTA) to calculate the heritability estimates on the brain structural variation. In order to capture the sparsity and localized correlation of genetic effect on the brain, we introduce a functional selection indicator via Bayesian nonparametric along with hierarchical Ising models to incorporate brain structural information. The efficacy of both methods is demonstrated in the data application and extensive simulations.