Hello - I have an issue that I would love some feedback on before I do something a bit too silly. We have this pipeline we have been working on

funcprofiler

for read based functional annotation of metagenomic profiles. A few of the tools in the pipeline are not in conda (eg

fmh-funcprofiler

or

mifaser

) so we have defined them as local modules with custom docker containers. Running the pipeline using the

singularity

or

docker

profiles seems to work well. The issue then comes with conda profile support for these steps. Do we a) not support conda for these steps b) try to do something potentially sketchy like adding some conditional logic to download the software from github if that step is called using the conda or mamba profiles (which would require further complications to make sure it supports both linux and windows etc) (and also feels a bit sketchy) c) require that the user install the software themselves and pass a path to the software location if they are using conda or mamba profiles? d) make an effort to add it to conda (despite not being the authors which also feels sketchy)? All seem less than ideal, but I suspect someone must have run into this problem before? 🤔

we usually recommend to do

d)

, and add them to bioconda yourself. It is basically the same as you adding someone else's software as an nf-core module 🙂

Thanks, that does seem like the best of the options listed. I think doing d could work for some of these tools, I am willing to give it a go. 🙂

and they are also not on pypi (just going of your comment about

pip

)?

They are not 😞

then bioconda it is 🙂

I stand corrected actually - one actually is on pypi! 🙂 So an easier path there at least! 🎉