Hi all! I’ve noticed a lot of stalled efforts to ...
# taxprofiler
n
Hi all! I’ve noticed a lot of stalled efforts to get Humann included as part of a nf-core pipeline. Can I make an argument that I think its natural home would be here in taxprofiler? My reasons are as follows: • The output from Metaphlan is one of the inputs for Humann (if you omit this, Humann has to rerun Metaphlan) • The Humann/Metaphlan databases and dependencies are quite interconnected already • under the hood, the first step is using a taxon-sprecific pangenome for annotation. Philosophically that makes me think it wouldn’t be out of place in taxprofiler 🙂 • as has been mentioned previously, there aren’t many read-based functional competitors — No need for a function-only read-based pipeline at this point • funcscan only focuses on contigs not reads; it doesn’t really have anything to do with MAGs so would be similarly out of place there.
j
I'm still not personally convinced. My impression is even though a few people have requested there haven't been anyone willing to actually develop a pipeline (I've advertised in different places as a 'free idea', but no-one has bitten), which makes me wonder how in demand it is. That said in one of the places where I talked about it, I was pointed to this pipeline: https://github.com/grp-bork/nHUMAnN It's not nf-core but from a good group, so I think this would suffice until someone comes up with other competitors to HUMANn3. I recognise the argument about the complementariness of MetaPhLAn --> HUMANn, I don't think we should break our clearly defined 'scope' (which is one of the things I'm most proud about taxprofiler 😅 - most metagenomics pipelines across bioinformatics so often result in feature creep and end up trying to do everything...). taxprofiler is not meant to be a generalised short-read metagenomics pipeline - but very specifically for generating taxonomic profiles - which HUMANn output is not. In otherwords, it's not about the input data but the output data.
👍 1
But this is my personal (very opinionated) opinion. @Lili Andersson-Li @Sofia Stamouli @moritz etc, as well as the wider community may have different opinions, so everyone can still chime in
l
I agree with James that including HUMAnN is out of taxprofiler scope😊
n
@James Fellows Yates @Lili Andersson-Li Thanks for considering it regardless! I’m then considering a read-based functional profiling pipeline to run humann, fmh-funprofiler, mifaser, RGI. In the interest of not reimplementing preprocessing and avoiding featyre creep, this pipeline will specifically expect the cleaned reads and the metaphlan profile from Taxprofiler. Any others that people want, before I head over to #C070369GP7T?
🙌 1
j
I think it's a good proposal to put forward to the community! In some ways it might be people don't want to run taxprofiler to prepare stuff because they want to go straight to the functional stuff, but we can discuss that in the proposal discussion on Https://GitHub.com/nf-core/proprosals
When you make it :)
n
Thanks, let the deliberations commence! https://github.com/nf-core/proposals/issues/59
🎉 1