OTUs or ASVs for 18S

[Alexandra_Bastkowska]

Hi all,

I want to characterize protists using 18S targeted sequencing. For taxonomic classification I want to use PR2. I tried to find standard approaches for taxonomically profiling 18s data but didn't find anything. Is it better to use OTUs over ASV for 18S data becuase of the larger variability of 18S? If otus, which algorithm is recommended, deblur or usearch or even something else?

Best wishes,
Alex

[SoilRotifer (Mike Robeson)]

HI @Alexandra_Bastkowska,

This is very much a question that is still discussed in the field. See the following articles

1. This paper from Callahan et al. 2017 is a great read.
2. Then you have Glassman & Martiny 2018 , which show that, broad scale patterns are similarly observed, between OTUs and ASVs.
3. This paper from Rob Edgar 2018 (he is the developer of usearch) highlights that clustering less, i.e. at 99% might be optimal.
4. Schloss 2021 warns about the potential splitting of genomes with ASVs.
5. Then you have one of my most favorite examples of the ASV approach from Debelius et al. 2020. In my view, this article highlights how ASVs can uncover what you might not notice when clustering into OTUs

In other words, I must respond with the dreaded it depends!

I think the choice depends on what you're after. How much resolution is required to address your research questions? Perhaps it does not matter which way you go. I've processed enough different datasets to know that, in some cases, I do not see differences between ASVs and OTUs. But with some other projects I see substantial differences between these approaches. This has to do in part with the fact that every environment, moment of sampling, etc is different.

Luckily QIIME 2 makes it quite easy for you to test both approaches. After you construct your ASV feature-table you can cluster that ASV feature-table into an OTU table, to your preferred level, i.e. 99 %, 98 %, etc... That is you can follow the instructions here, where you'd use your ASV table and ASV sequences as input into de novo clustering.

Personally, regardless of the marker gene / amplicon I prefer to proceed with ASVs. But again your mileage may vary.

Re OTU vs ASV: tl;dr technically speaking an ASV is an OTU. That is, the "Taxonomic Unit" is "Operational" given your current definition. That is, if you want to define your Taxonomic Unit of measure as a denoised sequence "clustered" at 100 % identity, then that is your OTU definition for the given study. In other words, ASV is a short hand for this definition.