After using DADA2, needs to take OTUs?

(Meha) #1

Dear Friends,
I would like to know after using DADA2 plugin, following by taking OTUs is necessary or clustering must be done? No need?

According to the cases I read I think it is not required, but I am in doubt:

  1. Reads should then be denoised into amplicon sequence variants (ASVs) or clustered into operational taxonomic units (OTUs).

  1. Algorithm (DADA), a model-based approach for correcting
    amplicon errors without constructing OTUs5. DADA identified
    fine-scale variation in 454-sequenced amplicon data while outputting few false positives


(Nicholas Bokulich) #2

Not needed. In addition to the links that you provided, you can also search this forum for many examples where this is discussed.

(Meha) #3

As I understood, I gained ASVs is instead of OTUs as a DADA2 output. In this regard, I need to know may I use q2-feature-table for my output or its methods are only used for OTUs?

By the way, for OTUs, three ways named de novo, closed reference, and open reference. In contrast, DADA2’s approach applys only reference sequences. My question: what is the meaning of this sentence written in the related-paper for DADA2?

" it was classified One Off"

If DADA2 does not identify a read that has not match with reference sequence how will the unidentified read be classified?

(Nicholas Bokulich) #4

Yes, you may.

I do not know — that question does not seem relevant to QIIME 2. You may want to ask the authors of that paper instead.

(Meha) #5

The paper is ralated to DADA2 and Qiime2 using it. The paper’s link is present in the Qiime2’s tutorial.

Now, so I ask it in another form. What is the behavior of Qiime2 against ASVs achieved by DADA2 which do not match with the reference sequence? How QIIME2 classifes the uncertain ASVs?

(Nicholas Bokulich) #6

Yes, I am aware. But that does not mean that everything written in the original article for dada2 is relevant to QIIME 2 or this forum. The passage you quoted relates to their benchmarking procedure, NOT to anything that dada2 actually does (let alone QIIME 2).

dada2 does not use reference sequences. The paper used reference sequences to test the specificity of the method. I recommend that you carefully re-read the original paper to understand how they are testing their method, and how that is distinct from what dada2 (and q2-dada2) actually does.