Data discretization for dynamic bayesian network based modeling of genetic networks

Authors:
Nguyen Xuan Vinh;Madhu Chetty;Ross Coppel;Pramod P. Wangikar
Affiliations:
Gippsland School of Information Technology, Monash University, Australia;Gippsland School of Information Technology, Monash University, Australia;Department of Microbiology, Monash University, Australia;Chemical Engineering Department, Indian Institute of Technology, Mumbai, India
Venue:
ICONIP'12 Proceedings of the 19th international conference on Neural Information Processing - Volume Part II
Year:
2012

Citing 7
Cited 0

Advances to Bayesian network inference for generating causal networks from observational biological data

Bioinformatics
A new dynamic Bayesian network (DBN) approach for identifying gene regulatory networks from time course microarray data

Bioinformatics
A Scoring Function for Learning Bayesian Networks based on Mutual Information and Conditional Independence Tests

The Journal of Machine Learning Research
BNFinder

Bioinformatics
Information Theoretic Measures for Clusterings Comparison: Variants, Properties, Normalization and Correction for Chance

The Journal of Machine Learning Research
Improvements in the reconstruction of time-varying gene regulatory networks

Bioinformatics
GlobalMIT

Bioinformatics

Quantified Score

Hi-index	0.00

Visualization

Abstract

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.