which version of cgat to use?

[mparker2]

I'd like to run the significant_bases_by_randomisation.py script. I tried installing iCLIPlib with cgatcore 0.5.4 and cgat-apps 0.5.3 installed from bioconda as they seemed to be the latest versions of CGAT, however a lot seems to have changed and so they aren't compatible with the iCLIPlib scripts. For example CGAT.Experiment.Start is now cgatcore.experiment.start (I assume) and CGAT.GTF is now cgat.GTF. What is the most compatible version of CGAT to use for iCLIPlib?

[mparker2]

I changed the module names to the lowercase versions and now it seems to run all the imports correctly so hopefully the differences are mostly superficial things like that... will see next week when I get the data to run :)

[IanSudbery]

Hi Matt,

SO the problem here is that iCLIPlib is written in python 2, and the lastest version of cgat are python 3 only. I've been meaning to port iCLIPlib over to python 3 for ages (like 18 months), but not gotten around to it. I'd be very interested to hear if you are running it in python 3 and its working okay.

[mparker2]

Hi Ian,
thanks for the response. I think I've got parts of iCLIPlib py3 compatible (at least enough to run the significant_bases_by_randomisation.py script) in this fork. There are probably other places that are still not py3 in the other scripts, though. I'd like to use the iCLIP_bam2geneprofile.py script as well so I'll see if that runs next...

I also added relative imports to the modules (not sure if that is a python 3 thing or not but I was getting errors) and added the scripts to setup.py.

[mparker2]

One thing I noticed is that I get a lot of infinite values for the FDRs. Is that expected behaviour due to not enough permutations?