Alzheimer’s Disease (AD) is a neurodegenerative and firmly incurable disease, and the total number of AD patients is predicted to be 13.8 million by 2050. Our motivation comes from needs to unravel a missing link between AD and biomedical information for a better understanding of AD. With the advent of data acquisition techniques, we could obtain more biomedical data with a massive and complex structure. Classical statistical models, however, often fail to address the unique structures, which hinders rigorous analysis. A fundamental question this dissertation is asking is how to use the data in a better way. Bayesian methods for high-dimensional data have been successfully employed by using novel priors, MCMC algorithms, and hierarchical modeling. This dissertation proposes novel Bayesian approaches to address statistical challenges arising in biomedical data including brain imaging and genetic data. The first and second projects aim to quantify effects of hippocampal morphology and genetic variants on the time to conversion to AD within mild cognitive impairment (MCI) patients. We propose Bayesian survival models with functional/high-dimensional covariates. The third project discusses a Bayesian matrix decomposition method applicable to brain functional connectivity. It facilitates estimation of clinical covariates, the examination of whether functional connectivity is different among normal, MCI, and AD subjects.