Adding ONT read support for ampliseq

[ShareiFae]

Description of feature

Would it be possible to add support for Oxford Nanopore Technology reads? I've found two projects (the second integrating the first) that might allow a quick implementation.

https://github.com/treangenlab/emu
https://github.com/genomic-medicine-sweden/gms_16S

[d4straub]

Hi @ShareiFae ,
emu seems indeed interesting, its also available as container https://bioconda.github.io/recipes/emu/README.html.
However, if I understood correctly, it compares sequences to a database and makes count tables from that. It is not a reference-free (de-novo) representative sequence (OTU/ASV/whatever) calculation like DADA2 or Deblur. Therefore I have the impression the tool is misplaced in this pipeline. Let me know if you have a different opinion.

[ShareiFae]

A shame, the downstream analysis and the report of this pipeline is really really good and I would like it to have it with nanopore data even it they cannot be done exactly the same way.

With EMU, as demonstrated by this other pipeline, you can create a phyloseq object that functions as Pseudo OTU tables:
https://github.com/josephpetrone/RESCUE/blob/master/detailed_steps.md#rstudio-and-phyloseq-handoff
https://github.com/josephpetrone/RESCUE

I've also found a pipeline that utilizes Quiime:
https://github.com/MaestSi/MetONTIIME

[samuell]

Hi @d4straub , I'm interested in hearing more about this. Is it the fact that it brings a database at all that you see as a problem here, or the size of the database?

EMU ships with a bundled database as an 84MB fasta file as part of the repo, but allows customizing this database, if that clarifies anything.

[d4straub]

Hi there,
databases are no problem. The "problem" is that many parts of the pipeline use ASV (/OTU) sequences. And EMU doesnt support that, is that correct?
There was/is also a discussion of nanopore analysis in nf-core slack: https://nfcore.slack.com/archives/CEA7TBJGJ/p1707383491473489
Essentially, newer nanopore data (from recent flowcells with high quality) might be possible to process with the pipeline and DADA2 with minor modifications (see benjjneb/dada2#759). Also possible is clustering with VSEARCH (that would be relatively easy to add into the pipeline).
Generally it is not only necessary to identify a solution, but also implementing it. I have neither the need nor time to do so. But if someone want to tackle that, I'm supportive.