Integration of genetic studies for multiple diseases has been shown to be a powerful approach to improving the identification of risk genetic variants. Although it has been shown that leveraging shared genetic basis among diseases, namely pleiotropy, can increase statistical power to identify risk variants, it still remains challenging to effectively integrative genome-wide association studies (GWAS) datasets for multiple diseases. In this presentation, I will discuss our novel DDNet-graph-GPA framework which addresses this challenge. graph-GPA is a novel Bayesian approach to integrate multiple GWAS datasets using a latent Markov random field architecture, which also allows incorporation of external prior knowledge. We also developed DDNet, a web interface where users can download a disease-disease graph obtained from a text mining of biomedical literature, and this disease-disease graph can be used as prior information for graph-GPA. The proposed approach improves the identification of risk variants and facilitates understanding of genetic relationship among complex diseases. I will illustrate the proposed approach with simulation studies and its application to real GWAS datasets.