Error tolerant model for incorporating biological knowledge with expression data in estimating gene networks

We propose a novel statistical method for estimating gene networks based on microarray gene expression data together with information from biological knowledge databases. Although a large amount of gene regulation information has already been stored in some biological databases, there are still errors and missing facts due to experimental problems and human errors. Therefore, we cannot blindly use them for understanding gene regulation and a robust procedure with a statistical model for using such database information is required. By using gene expression data, we provide a probabilistic framework of a joint learning model for repairing database information and for estimating a gene network based on dynamic Bayesian networks, simultaneously. To show the effectiveness of the proposed method, we analyze Saccharomyces cerevisiae cell-cycle gene expression data together with KEGG information.