<http://biology.plosjournals.org/perlserv/?request=get-document&doi=10.1371/journal.pbio.0050008> Interesting, I think we should test Lemone on this data set ASAP. They report 80% precision for their algorithm, but at 4% recall, that's F-measure 0.076 which is not that impressive (we get 0.22 on simulated data). I wonder why they didn't include Genomica in their comparison. Was it too good? Or was it because of this: "None of the publicly available Bayesian network learning software that we tried were designed to infer a network of the size we attempted with this study, nor would they run to completion. Those that did not crash due to memory problems were unable to generate a network after several weeks of computational time." A bit hard to believe for a network of some 4000 genes. Maybe we can write a comment if Lemone performs much better than whatever Bayesian network algorithm they tested? Tom