Mining the genome and regulatory networks

The conference on genome informatics held in Yokohama at the end of last year provided an international forum for disseminating the latest developments and applications in advanced computational methods that can be used for solving several biological problems. In bioinformatics and systems biology the topics covered included the prediction of regulatory networks, microarray data analysis, and graph mining for structural biological data.Biological processes in microbial cells are known to be carried out through interactions between multiple 'functional modules' that serve as the basic building blocks of complex biological networks. At the gene level, for example, a module is defined as a set of genes that can be grouped according to their biological function in a pathway or process, and in many cases modular structure and module components are highly conserved. Ying Xu (University of Georgia, Atlanta, USA) presented a computational method for predicting modules in microbial genomes that is based on the fact that neighboring genes in prokaryotic genomes are likely to be functionally related, as a result of the operon organization. Using 224 microbial genomes from 175 different species, Xu and colleagues quantified the functional relatedness among genes on the basis of their presence (or absence) and relative proximity in different genomes, and obtained a gene network in which all possible pairs of genes had a score representing functional relatedness. By applying a threshold-based clustering algorithm to this network, numerous functional modules were obtained. The predicted modules were statistically and biologically significant, and the genes of a module were more likely to share Gene Ontology 'biological process' terms than terms relating to 'molecular function' or 'cellular component'. These predictions could be used to guide experimental designs for investigating particular biological processes and might also provide a basis for further prediction of detailed gene