|
Abstract:
|
Identification of genetic risk factors based on single genes has turned out to have high false positive rate in the studies of genome-wide association. Meanwhile, it is widely acknowledged that diseases are typically triggered by the confluence of multiple genetic and external factors. We develop a Bayesian model for deciphering genotype-phenotype mapping by integrating high dimensional, correlated, and mixed data. The model relaxes the assumption of conditional independence, and the joint association can be derived between two groups of variables. The model features fast computation by optimizing the model structure and the posterior calculation, and it is applicable to categorical or continuous data. We test the proposed methods using the simulation studies. We demonstrate the advantages using the genome-wide association studies that have difficulty in deciphering complex disease-causing mechanism and genetic risk factors.
|
