Probabilistic graphical modelling of causal effects among the occurrences of transcription factors in DNA sequence

Abstract

Genome mapping of transcription factor targeted by ChIP jointly with microarrays or sequencing procedures is a powerful instrument for laying a foundation for understanding transcriptional regulatory networks. Hence the need for computational methods that can form the basis of experimental verification of these networks. We employ a probabilistic graphical model of the form of linear Gaussian Bayesian network to model causal effects between transcriptional factors (TFs) in two genome datasets. The bnlearn R statistical package is used for learning the network structure of the ENCODE pilot data and Mouse Embryonic Stem Cell data. Our results show that the Bayesian network efficiently model the causal effects between TFs, handle uncertainty with respect to probability theory and establish indirect with direct causation. Finally, an integrated Bayesian network model