Open djinnome opened 4 years ago
Hi Jeremie,
Thanks for your interest in our tools and repository. Just to clarify a bit, this is CARNIVAL. COSMOS is the Causal integration of multi-omics data with prior knowledge. In this repo, we are only using transcriptomics data and a protein-protein interaction network to run CARNIVAL. In the COSMOS publication, transcriptomics, metabolomics and phosphoproteomics are integrated in a more complex causal network also containing, among others, metabolic reactions. Then CARNIVAL is applied on these more complex networks in COSMOS.
I strongly recommend you to use cplex to run CARNIVAL. We included the possibility of using the cbc and lpSolve solvers because they are totally free. However, we show that the performance of cplex, mainly in large networks, is significantly better. Cplex has a free version for academia. In case you want to keep working with cbc, you must increase the timelimit
parameter. The error you are getting means that cbc was not able to find a "feasible" solution in the specified timelimit
.
Best regards, Alberto.
Hi @alberto-valdeolivas Thanks for the clarification! I would love to use cplex, but IBM does not consider a national lab to be an academic institution. I then tried CBC with longer time limits, but to no avail.
Fortunately, the folks at Gurobi granted me a license as long as I use it for academic work. I have successfully run gurobi on the LP file that CARNIVAL generates (although the 255 character limit in gurobi required me to rename a couple of the variables), and then I went ahead and extended CARNIVAL with gurobi functionality. Now if solver="gurobi"
then it will create a gurobi Command file with the appropriate gurobi parameters, and it will generate a result file similar to CBC output that is parsed by exportResultGurobi.R
. I went ahead and created a pull request, but before you accept, I wanted to discuss the possibility of a general fix for the 255 character limit on the variables that gurobi chokes on.
Hi again Jeremy,
Many thanks! I will discuss with the person in charge of the CARNIVAL package maintenance in order to add this new solver.
Best, Alberto.
Hi folks,
Thanks for providing this example of how to apply COSMOS to viral pathogenesis data. I am trying to reproduce your pipeline, and I got as far as runCarnival, but the problem now is that I don't have access to CPLEX, so I installed lpSolve and CBC, but neither solver completes. This is the error I get with CBC: