Examining the Suitability of Dynamic Bayesian Networks for Discovering Genetic Regulatory Interactions R.J.P. van Berlo E.P. van Someren M.J.T. Reinders Information and Communication Theory Group,

Learning regulatory interactions between genes from microarray measurements presents one of the major challenges in functional genomics. Thus far, the use of Bayesian networks to learn these interactions has been primarily restricted to static pertubation experiments. This paper studies the suitability of learning dynamic Bayesian networks when observing a dynamic response to a particular perturbation. Through extensive artificial-data experiments it is investigated how the performance of discovering the true interactions depends on varying data conditions. These experiments show that the performance most strongly deteriorates when the connectivity of the original network increases and that more than a proportional increase in the number of samples is needed to compensate for this. Furthermore, it was found that a lower performance is achieved when the original network size becomes larger, but that this decrease can be greatly reduced with increased computational effort. The method is illustrated on the Alpha subset of the S. cerevisiae gene expression data of Spellman.