Hi Steven, I think the problem is not so difficult. For the regulatory network, after selecting the most significant edges, I used to define modules as sets of genes having the same set of regulators. The only problem is that sometimes (especially for genes with just 1 regulator) this mixes genes that are activated with genes that are repressed by the same regulator, and the figures look a bit weird. So I would redefine modules to be sets of genes with the same set of regulators with the same regulation sign. The latest output prints for each edge the regulation sign as well as the conditions in which it is 'active' (I wrote this on the fly while in Leuven last Friday so it's not yet in the repository ... :-), and I will incorporate this for drawing figures sometime this week. The papers you attached are interesting, they seem like alternatives for the fuzzy clustering method we used in our Gibbs sampler paper, right? Tom On 31 Aug 2007, at 16:44, Steven Maere wrote:
Hi all,
I've been thinking a bit about the problem of having multiple clustering solutions and combining these into 'consensus modules'. I don't know how much and along which lines you have been working on this problem, Tom ? My first thought was to somehow cluster the weighted summary graphs that Tom and Anagha produce with their software. Weighted bipartite graph clustering algorithms might be ideally suited for this, and there are apparently some algorithms and implementations available. I've attached a few relevant articles. We could think about: - clustering the weighted bipartite regulator-gene graph produced by the more recent Lemone 'S' versions - clustering similarly constructed bipartite regulator-condition graphs, as Eric suggested, to see which conditions matter for which regulators
Let's discuss further after Tom gets back from holidays.
S
-- ================================================================== Steven Maere, PhD
Dept. of Plant Systems Biology Bioinformatics & Evolutionary Genomics research group VIB / Ghent University Technologiepark 927 B-9052 Gent BELGIUM tel: +32 9 331 3810 fax: +32 9 331 3809 email: stmae@psb.UGent.be http://www.psb.UGent.be http://bioinformatics.psb.ugent.be/ ==================================================================
<ahma07.pdf> <fern04.pdf> <zha01.pdf> _______________________________________________ Beg-sysbiol mailing list Beg-sysbiol@psb.ugent.be https://maillist.psb.ugent.be/mailman/listinfo/beg-sysbiol
-- Tom Michoel <http://www.psb.ugent.be/~tomic/>