Hi LeMoNe-people, As you know, I talked with Aviv about the Arabidopsis project I'm working on. I also explained the (latest) version of the LeMoNe algorithm that I'm using - running Gibbs sampling several times to find different gene and experiment clustering results, creating tight clusters from these results, and then learn regulatory programs for these tight clusters. Aviv really liked this approach a lot, especially since it intends to tackle the weak points of the Geronemo approach such as creating robust clusters (allowing genes to be filtered out) and improving computational efficiency. She suggested that this LeMoNe approach might further be improved by running the full original LeNoNe version (iterative optimization of both learning the regulatory programs for each module and gene reassignment to the regulatory program that best explains the particular gene's expression behaviour) on the tight clusters, so not just learning the regulatory programs on the tight clusters, but also allowing genes to be reassigned to their best regulatory program, and then to optimize both iteratively. This adaptation will likely decrease LeMoNe's computational efficiency but might lead to better results. I tought this suggestion might be of interest to all LeMoNe people? See you tomorrow! Kind regards, Nathalie