Functional prediction of B. subtilis genes from their regulatory sequences

In bacterial cells, gene expression is regulated by multiple sigma factors, each of which has its promoter specificity, according to their conditions. Thus, if we can discriminate which sigma factor binds to the upstream region of a given coding sequence, we can predict in what condition it will be expressed. In this paper, we show this approach is feasible for the analysis of Bacillus subtilis genome. Based on our collection of known promoter sequences, we prepared 8 predictors to characterize known sigma factors using the hidden Markov model and their prediction accuracies were estimated with a cross-validation test. Furthermore, we predicted the sigma-dependencies for each of 1415 candidate genes in the genome. Our prediction results are experimentally testable and seem useful for the post-sequencing project.