nanoseq
nanoseq copied to clipboard
nf-core
Add guppy to BioConda
I cant seem to find GraphMap2 on BioConda which also means the associated BioContainer wont exist. It would be nice to be able to use this in the pipeline because it has some updated functionality for Nanopore protocols.
Copying in from Slack discussion on this:
....
https://github.com/lbcb-sci/graphmap2 Which is version 0.3.0 and packaged in bioconda as graphmap https://bioconda.github.io/recipes/graphmap/README.html
a bit odd since there aren't many releases and the bioconda packaged one is from the author Ivan Sovic's repository that instead links to the new one in his (former?) group...
Also odd that the new repository doesn't have the > 0.3X releases on it - so probably should ping the authors and then make a new release with the new URL? Could also ask over at Bioconda gitter with respect to this?
Requested by @dpryan79 here https://github.com/lbcb-sci/graphmap2/issues/6
The BioConda GraphMap version has been updated to the latest release
https://bioconda.github.io/recipes/graphmap/README.html
ok, then adding in - bioconda::graphmap2=0.6.3 to the environment.yml and you should be fine 👍
Except we aren't using conda because guppy isn't available there and could be problematic adding the gpu version :sweat_smile: .Using a mix of containers on DockerHub and BioContainers to overcome this.
Hmm...good question :+1: Would have to ask Nanopore. In theory, it should support both gpu and cpu out of the box but in practice I couldn't use the cpu container for gpu. Possibly some gpu libraries missing in the container. When I looked it was quite difficult getting information regarding the software and any docs - needed to have a Nanopore account.
I have a nanopore account, so I can ask them if you'd like (they're not terribly speedy at replying).
Docker containers we are currently using in the pipeline:
GPU Docker container: https://hub.docker.com/r/nanozoo/guppy_gpu/dockerfile
CPU Docker container: https://hub.docker.com/r/genomicpariscentre/guppy/dockerfile
Thanks @dpryan79 . That would be amazing! I'm actually supposed to be on paternity leave. Apparently 🤪
Ok, if I'm not mistaken there, one could simply work on adding in the required CUDA libraries to the container itself: nvidia/cuda:9.0-base
If you're running with Singularity, these are then used to accessing the GPU(s) automatically for you.
@drpatelh I'll post a query on their forum then. Don't spend too much time on github :)
For those with an ONT account I've asked about guppy redistribution: https://community.nanoporetech.com/posts/guppy-redistribution
So, the reply from ONT is that guppy cannot be redistributed at the moment. The docker containers are actually breaches of their license, which I suppose isn't an issue for nf-core (you're just using them, not providing them), but the people who made them could theoretically get sued. Granted, the odds of that are very low (especially since ONT is allegedly going to at least provide their own containers somewhere), but they're not 0. BTW, for those of us with clusters, we're actually supposed to restrict access to guppy to only ONT customers...not sure how that's really possible in practice without them having their own copies.
Thanks @dpryan79. So basically it can't be added to BioConda (unless the license is changed or we get explicit approval). Not sure what to say about the people hosting the containers on Dockerhub - I think they are actually doing the Bioinformatics community a service by hosting them before they are officially added but I guess it doesn't matter what I think. As far as I know, only people with an official account are using the pipeline at the Crick but this is difficult to guage once they are made available to download and run.
Yeah, ONT needs to just relax their licensing and make everyone's life easier.
Maybe something to consider as a successor to Guppy, apparently even better: https://github.com/nf-core/nanoseq/issues/61