Interpretation of genetic regulatory networks using protein interaction and transcription information

Abstract

Constructing entire genetic regulatory networks is an ambitious goal for understanding and representing living mechanisms of a cell. Although several large scale experiments have tried to infer a lot of relations between genes, most of the results do not represent direct causal relations of genes, such as transcriptional activation or repression, but a set of obscure relations which rely upon the statistical or structural probability of the applied learning algorithms. These obscure relations are believed to have a lot of false positives and hardly interpreted to biological meanings of the target cell. To interpret and extract useful and reliable information from large scale regulatory networks, we developed an automated causal relation screening method for genetic regulatory networks by modeling biologically possible flows of proteins; a viaduct. In the experiment on a Bayesian genetic regulatory network of yeast cell cycle, we have found 821 viaducts by constructing a partial directed graph which consists of protein-protein interaction data and transcription factorinformation of Saccharomyces cerevisiae. Finding and inspecting those viaducts provides a new way to extract uncovered causal relations from large scale genetic regulatory networks and to measure how much the networks are constructed to contain reliable information.