Data Discretization for Dynamic Bayesian Network Based Modeling of Genetic Networks
Dynamic Bayesian networks (DBN) are widely applied in Systems biology for modeling various biological networks, including gene regulatory networks and metabolic networks. The application of DBN models often requires data discretization. Although various discretization techniques exist, currently there is no consensus on which approach is most suitable. Popular discretization strategies within the bioinformatics community, such as interval and quantile discretization, are likely not optimal. In this paper, we propose a novel approach for data discretization for mutual information based learning of DBN. In this approach, the data are discretized so that the mutual information between parent and child nodes is maximized, subject to a suitable penalty put on the complexity of the discretization. A dynamic programming approach is used to find the optimal discretization threshold for each individual variable. Our approach iteratively learns both the network and the discretization scheme until a locally optimal solution is reached. Tests on real genetic networks confirm the effectiveness of the proposed method.
KeywordsDynamic Bayesian network Gene regulatory network Discretization Mutual information Microarray
Unable to display preview. Download preview PDF.
- 3.Friedman, N., Goldszmidt, M.: Discretizing Continuous Attributes while Learning Bayesian Networks. In: Proc. ICML, pp. 157–165. Morgan Kaufmann (1996)Google Scholar
- 6.Murphy, K., Mian, S.: Modelling Gene Expression Data Using Dynamic Bayesian Networks. Tech. rep., Computer Science Division, University of California, Berkeley, CA (1999)Google Scholar
- 7.Myers, J.L., Well, A.: Research Design and Statistical Analysis, vol. 1. Psychology Press (2003)Google Scholar
- 14.Yao, Y.Y.: Information-theoretic Measures for Knowledge Discovery and Data Mining. In: Karmeshu (ed.) Entropy Measures, Maximum Entropy Principle and Emerging Applications, pp. 115–136. Springer (2003)Google Scholar