A Search-Based Approach for Bayesian Inference of the T-cell Signaling Network

We apply a search-based technique for learning high-quality Bayesian networks to proteomic flowcytometry data for a portion of the human T-cell signaling network. Although Bayesian network models have been learned from this data using methods such as MCMC that sample from a posterior distribution [1], [2], [3], we demonstrate that our more comprehensible searchbased technique, which uses model averaging by Bayesian bootstrap replicates, provides comparable results. We also show that additional edges not in the consensus model, but identified by both our technique and that of Eaton and Murphy are very unlikely to be artifacts of the network learning technique, and should be investigated further.