brendelgroup / movrs Goto Github PK

We probably should create a good logfile that traces the consensus motif generation in this script. This would include untangling the final names - the current suffixes were meant for a bit of tracing during development, but this is inadequate for code distribution.

We also need to think about some edge cases of motif merging. In principle, it could be that, for example, we have two 8-mer motifs that overlap by a 6-mer, with 2-mer overhangs on each end: do we report the 6-mer as a consensus motif (which may be found everywhere) or the 10-mer (which may not be found anywhere) or else? The code is flexible enough to go any direction, but we need to review what it is doing currently and document choices.

Coding assigned to me, but issue formulation and example collection is an issue for the team.

I think it may be more conservative to uses "spaces" instead of "tabs" to indent lines in python code.
To wit, vi *pl seems to equate tab with 8 spaces, whereas vi *py sets tab to 4 spaces. Thus "tabs" appear definition-dependent and may cause problems when we edit the files with different editors.

No doubt this is all well known, and of course I know very little about python programming.

But it's worth following up, thus this "issue".

When running a test MOVRs job I noticed an error: AttributeError: module 'networkx' has no attribute 'connected_component_subgraphs.

It appears this error is due to connected_component_subgraphs being deprecated in networkx v. 2.4.

A note about this causing an error in another package is here.

To address this I have installed networkx v 2.3 (the last version where the definition is found) as follows:
pip install --user 'networkx==2.3'. I will run the same test with this version installed and report back.

If we don't want to specify a specific version of the networkx package, this solution may be appropriate for our case.

In Step 7, we need to document candidate MoVRs motif occurrence statistics in the validation sets. We derived motifs based on presence in some form (!) in some minimal number of the training sets - but how does that work out in the validation sets with the derived, consensus motifs? The thought of the whole workflow is that we have at this point solid motifs that indeed are genome-wide over-represented. We should test that.

Something Yao can explore - it's a simple enough addition to the MoVRs script, but testing and summary statistics will take a bit of effort.

I recently encountered an apparent error when running the sample test using the newest build of MoVRs:

MoVRs -a testpeakfile --genome ./TestGenome -o test_0906 --size [-60,40] -p 8 >& errTEST_0906

The error reads as follows:

python /home/rtraborn/development/MoVRs/scripts/MoVRs_ExtractMotif.py -i AllMotifs.meme -t 1e-10 -o FilteredMotifs.meme File "/home/rtraborn/development/MoVRs/scripts/MoVRs_ExtractMotif.py", line 27 tmp_return = [x[0] for x in tmp_sorted] ^ TabError: inconsistent use of tabs and spaces in indentation Fatal error running MoVRs_ExtractMotif.py. Please check.