Estimating gene networks from gene expression data by combining Bayesian network model with promoter element detection

Abstract

We present a statistical method for estimating gene networks and detecting promoter elements simultaneously. When estimating a network from gene expression data alone, a common problem is that the number of microarrays is limited compared to the number of variables in the network model, making accurate estimation a difficult task. Our method overcomes this problem by integrating the microarray gene expression data and the DNA sequence information into a Bayesian network model. The basic idea of our method is that, if a parent gene is a transcription factor, its children may share a consensus motif in their promoter regions of the DNA sequences. Our method detects consensus motifs based on the structure of the estimated network, then re-estimates the network using the result of the motif detection. We continue this iteration until the network becomes stable. To show the effectiveness of our method, we conducted Monte Carlo simulations and applied our method to Saccharomyces cerevisiae data as a real application.

Probabilistic Models for Gene Silencing Data

“…Instead of the parametric approaches discussed so far, the relationship between parents and children in the DAG can also be modeled by non-parametric regression models [64,65,66,134]. The result is a non-linear continuous model.…”

Microarray data analysis and mining approaches

“…Tools allowing the integration of microarray data with promoter structure information have been developed [104][105][106][107][108][109]. The software developed by Kel [104,106] is commercially available (Explain software, www.biobase.de) and uses a genetic algorithm to predict relevant promoters in a set of given transcripts obtained from microarray analysis, taking advantage of the promoter element matrix database TRANSFAC [110][111][112].…”

“…Werner software [107,108] is also a commercial tool where promoter elements are identified using MatInspector [113]. Tamada [109] instead has developed a statistical method for estimating gene networks and detecting promoter elements simultaneously. This method integrates microarray gene expression data and the DNA sequence information into a Bayesian network model.…”

Machine Learning in Bioinformatics

“…There are several works that use static Bayesian networks to model genetic networks [162,[165][166][167][168][169][170][171][172][173][174][175][176], In Tamada et al [177] DNA sequence information is mixed with microarray data in the Bayesian network in order to obtain a more accurate estimation of the network when the number of microarray data is limited. In Nariai et al [178] genetic networks estimated from expression data are refined using protein-protein interactions.…”