A Bayesian connectivity-based approach to constructing probabilistic gene regulatory networks

• DOI: 10.1093/bioinformatics/bth318
• Publication Date: 2004-05-18T00:43:56Z
• Authors: Xiaobo Zhou; Xiaodong Wang; Ranadip Pal; Ivan Ivanov; Michael Bittner; Edward R. Dougherty
• Keywords: Gene regulatory network; Dynamic Bayesian network

Abstract Motivation: We have hypothesized that the construction of transcriptional regulatory networks using a method optimizes connectivity would lead to regulation consistent with biological expectations. A key expectation is hypothetical should produce few, very strong attractors, highly similar original observations, mimicking state stability and determinism. Another central that, since it expected control distributed mutually reinforcing, interpretation observations small number connection schemes. Results: propose fully Bayesian approach constructing probabilistic gene (PGRNs) emphasizes network topology. The computes possible parent sets each gene, corresponding predictors associated probabilities based on nonlinear perceptron model, reversible jump Markov chain Monte Carlo (MCMC) technique, an MCMC employed search configurations find those highest scores construct PGRN. has been used PGRN observed behavior set genes whose expression patterns vary across melanoma samples exhibiting two different phenotypes respect cell motility invasiveness. Key features faithfully reflected in model. Its steady-state distribution contains attractors are either identical or states data, many singletons, which mimics propensity stably occupy given state. Most interestingly, rules for most optimal generated constituting remarkably similar, as be operating basis, interactions between components. Availability: appendix available at http://gspsnap.tamu.edu/gspweb/pgrn/bayes.html username: gspweb password: gsplab. Supplementary Information: