as far as I know MCL does not support prioritizing seed nodes or giving seed node weights
Dear all, Which are the well used protein protein interaction clustering methods that uses weights/scores given to nodes for module detection ? I know about jActiveModules in Cytoscape and BioNet R package. I don't prefer to use jActiveModules (because of random seed selection for each run, non-repeatable results), and right now I am playing around with BioNet package. I did not come across any other node weighted clusteing methods do you know about any of such algorithms used for ppi interaction clustering or graph clustering in general ?
Edit : I also found out about MINE cytoscape plugin, a perl implementaion of the same plugin takes node weights as input, but the algorithm just uses these scores as a threshold to remove which ever node that doesn't have minimum input score.
Thanks a lot.
Edit 2: I got a chance to talk with a Professor in algorithmic bioinformatics yesterday and what he told me was that what I want is actually called Prize-Collecting Steiner Tree Problem and he suggested using the R Bionet package
1 answer
Try MCL algorithm. It is widespread among biologists.
Log in to answer this question.
Can you specify the random seed for jActiveModules? Most programs that need a random seed allow for one to be specified in order to ensure reproducibility. I would be very surprised if you couldn't do this.
In a way yes, in jActiveModules menu there is an option "Non-Random Starting graph", but my experience is that using this option, the module clustering will be identical for that session, but when I re-do analysis after restarting cytoscape and opening the same session, module clustering changes.