Add VirSorter2 process as an alternative to VirSorter #128
Conversation
|
Great, thanks a lot! @mberacochea @guille0387: @fischer-hub is working with me at the RKI and we anyway wanted to use VS2 for some annotations. Thus, David kindly added the functionality to VIRify. Of couse, this would need more proper benchmarking but the idea would be to swap VS against VS2 at some point. For now, there would be this parameter switch possible. |
|
@hoelzer @fischer-hub this is amazing, thank you so much. This has been in our backlog for ages. I will go over the PR as soon as I can, but from a quick read it looks great. |
There was a problem hiding this comment.
Thanks @fischer-hub, great stuff. I left a few comments, nothing major.
The one thing I would like to add is a few more unit tests for parse_viral_pred.py, can you send me an example of the virsorter2 outputs?
virify.nf
Outdated
| viphos_annotations = annotation.out.map { _, __, annotations -> tuple(annotations, i++) }.collect(){it -> it[0]} //{ annotations, count -> "$annotations".replace('.', '_' + count + '.') } | ||
| taxonomy_annotations = assign.out.map { _, __, taxonomy -> tuple(taxonomy, j++) }.collect(){it -> it[0]} //{ taxonomy, count -> "$taxonomy".replace('.', '_' + count + '.') } | ||
| checkv_results = checkV.out.map { _, __, quality_summary, ___ -> tuple(quality_summary, k++) }.collect(){it -> it[0]} //{ quality_summary, count -> "$quality_summary".replace('.', '_' + count + '.') } |
There was a problem hiding this comment.
This is to support the -list of assemblies, correct?
There was a problem hiding this comment.
When running on a list of samples this was my first fix to prevent the write_gff file collision, but I think this should be sufficient as well since then the input files are enumerated and have different names too. (I just forgot to remove this afterwards, will check again)
Also this might be solved when removing the .first() from contigs.first() later on because then each sample will run in its own write_gff process instance? Or could there be annotation files from the same sample with the same name too, e.g.: from different contigs of the same assembly?
There was a problem hiding this comment.
I see. The change on the contigs.first() it a but more invovled as we need to .join() the assemblies with the corresponding annotations. Otherwise, the pipeline could mix results from different assemblies.
There was a problem hiding this comment.
Yes, true. Maybe its better then to keep the fix and handle the bug in a separate PR.
Sure, will go over the changes asap and attach some example output here! |
Co-authored-by: Martín Beracochea <mbc@ebi.ac.uk>
Co-authored-by: Martín Beracochea <mbc@ebi.ac.uk>
Co-authored-by: Martín Beracochea <mbc@ebi.ac.uk>
Added VirSorter2 as a new process that can be used instead of
VirSorterusing the new flag--use_virsort2, adapted the downstream processes (theparseprocess and GFF generation script) to work with the different results ofVirSorter2. E.g.:VirSorter2reports a confidence score (0-1) for every viral hit in the input data instead of a category.Also had to add some changes to the GFF generation process because I always ran into file collision issues when more than one of the input samples reported significant viral sequences, since the
VirSortandVirFinderreports have the same file name for every sample. Might also be solved by running the GFF script for every sample containing viral seqs instead of once for all samples but I already asked about this in #127, maybe I'm missing something here!