Figure 1.

Inference of a static Bayesian GRN. Bayesian GRNs were generated from two microarray datasets (1) time course of primary ECs in conditions of SFD for 24 hours (8 time points in triplicate) to induce apoptosis and (2) disruptant dataset generated from the siRNA-mediated knockdown of 351 transcripts. These two datasets were used in network inference. Bayesian GRNs were generated to maximise the posterior probability, which consists of two priors; (a) the dynamic Bayesian GRN prior (generated from the time course data) and (b) the array prior (measuring the relationships between the gene knockdowns and their regulatees, as measured by z-score in the 351 disruptant dataset), as well as the marginal likelihood. This is the non-parametric regression through estimated edges based on the 351-disruptant dataset. The gene list of 694 transcripts chosen for network inference was selected based on (1) the transcripts regulated during the apoptosis time course and (2) the 351 siRNA targeted transcripts. Using the dynamic Bayesian GRN as a prior for the disruptant dataset, the relationships for the 694 transcripts within the 351 disruptant dataset were inferred. Bootstrapping of the network prior and the estimated static network helped improve edge reliability in the final network. The static apoptosis Bayesian GRN can be viewed using Cell Illustrator, which can be freely downloaded from webcite.

Affara et al. BMC Genomics 2013 14:23   doi:10.1186/1471-2164-14-23
Download authors' original image