A Bayesian Algorithm for Reconstructing Two-Component Signaling Networks
We present an algorithm, based on a Bayesian network model, for ab initio prediction of signaling interactions in bacterial two-component systems. The algorithm uses a large training set of known interacting kinase/receiver pairs to build a probabilistic model of dependency between the amino acid sequences of the two proteins and uses this model to predict which pairs interact. We show that the algorithm can accurately reconstruct cognate kinase/receiver pairs across all sequenced bacteria. We also present predictions of interacting orphan kinase/receiver pairs in the bacterium Caulobacter crescentus and show that these significantly overlap with experimentally observed interactions.
KeywordsHistidine Kinase Domain Architecture Receiver Domain Bayesian Network Model Bayesian Algorithm
Unable to display preview. Download preview PDF.