Several human diseases are thought to have complex genetic causes, where mutations of several genes are necessary to lead to the occurrence of the disease. Unfortunately, the affected genes are difficult to detect using clinical studies alone. I will discuss an approach we are developing at the National Center for Integrative Biomedical Informatics (U. Michigan) that allows us to generate possible molecular interactions that could explain the clinical results. We will look especially at Bipolar Disorder as a motivating example.