Using bayesian networks to learn gene regulatory networks

We present a variant of the Fast Greedy Equivalence Search algorithm that can be used to learn a Bayesian network that represents the full transcriptional regulatory network of the human brain. We have fully implemented the algorithm and have some preliminary results that show that we can retrieve the Markov Blanket of individual genes of interest with good accuracy and reasonable time (2 hours in a 12 core 2.6GHz computer). The algorithm is parallel and we are currently waiting to deploy it in a supercomputer where we expect to be able to learn the full genome network. This network could then be used as an exploratory tool by the biology community when studying the relationships between genes