Bayesian Network Analysis Software: Interfacing Visualization and User Interaction

A challenge in computational biology is to uncover the gene-protein inter-actions from gene expression data. A good approach for discovering inter-actions between genes based on multiple expression measurements is to use Bayesian networks. To that end, BUBBLE (Bayesian User Based Bio-logical Learning Environment) has been developed. BUBBLE allows a re-searcher to create pathways, to specify prior constraints, to import data from various databases and learn a Bayesian network model. In addition, User Relevance Feedback queries the user with optimal chosen questions to improve the correctness of the pathways. The goal of the project was to develop an appropriate and intuitive environment so that the user could be effective at contributing to the Bayesian network construction and analysis.