Redes bayesianas para inferÃncia de redes regulatÃrias de genes

AUTOR(ES)
DATA DE PUBLICAÇÃO

2005

RESUMO

With the development of functional genomics, data on a great number of species are being obtained in huge volumes. Technologies to measure the differences of the gene expression, through mRNA concentration (microarray), have become extremely popular, and their costs are decreasing. The reconstruction of genetic networks from gene expression data to study the organism dynamics is an important process and raises the challenge of connecting genes and their products in metabolic pathways, circuits and functional networks. Understanding gene regulatory networks can provide valuable information for treatment of diseases, identification of genes that control and regulate cell events, and discovery of complex metabolic pathways. A genetic regulatory network is a model that represents the regulations between genes using a directed graph where the nodes indicate genes, and an edge (Gene 1, Gene 2) indicates that Gene 1 regulates Gene 2 (with activation and/or repression). Several methods have been proposed during the last years to infer a genetic network from microarray data using mathematical models, such as differential equations, Boolean networks, and Bayesian networks. In the present work we show the use of a Bayesian network model for inferring genetic networks from microarray data. Two different programs were implemented: one using a Bayesian network model and another one using a dynamic Bayesian network model, both with non-parametric regression. We use Bayesian Information Criterion (BIC), a simpler but still effective approach, to choose the best networks. Our results were compared to those of previous works, using two datasets: an artificial dataset to infer an artificial gene regulatory network; and gene expression microarray data of Saccharomyces cerevisiae to infer the TCA cycle (tricarboxylic acid). Experiments with artificial data produced good results comparing to previous models, mainly when prior information was added. The experiments with gene expression data were more surprising, as even though only a small sample was available, results were as good as those found by previous models. Regulatory gene networks inference from microarray data is a recent and difficult problem. This work presents a simpler model which obtained promising results, and that may be extended in future works

ASSUNTO(S)

inferÃncia de redes regulatÃrias de genes non-parametric regression otimizaÃÃo bayesian networks rede bayesiana regressÃo nÃo-paramÃtrica ciencia da computacao optimization inference of gene regulatory networks

ACESSO AO ARTIGO

Documentos Relacionados