Gene network prediction from microarray data by association rule and dynamic Bayesian network

Abstract

Using microarray technology to predict gene function has become important in research. However, microarray data are complicated and require a powerful systematic method to handle these data. Many scholars use clustering algorithms to analyze microarray data, but these algorithms can find only the same expression mode, not the transcriptional relation between genes. Moreover, most traditional approaches involve all-against-all comparisons that are time consuming. To reduce the comparison time and find more relations, a proposed method is to use an a priori algorithm to filter possible related genes first, which can reduce number of candidate genes, and then apply a dynamic Bayesian network to find the gene's interaction. Unlike the previous techniques, this method not only reduces the comparison complexity but also reveals more mutual interaction among genes. © Springer-Verlag Berlin Heidelberg 2005.