http://www.biomedcentral.com/1471-2105/12/89
The main goal is “to develop a method that can discover SNPs associated with disease from real data”.
What I like in the article is that the authors try to find and justify an appropriate criteria to how to choose the most correct network.
What is annoying: they were not able to answer the question unambiguously

Have you ever tried to use Bayesian network for learning epistasis? What are your impressions? Can you give further interesting links for BN - epistasis?