Getting a bit lost in the workflow - post deblur

[mstagliamonte]

Dear Qiimers,

I am moving my first steps with Qiime 2 on 16S data, but I am a bit lost when moving from denoising the data with deblur to clustering (using vsearch?)

Here is my workflow so far:

1. Import demultiplexed fastq files (illumina PE)

2. Remove remenant adapter contamination (qiime cutadapt trim-paired)

3. Join pairs (qiime vsearch join-pairs)

4. Quality filtering (qiime quality-filter q-score-joined)

5. Denoising using deblur (qiime deblur denoise-16S)

So far, so good. Now I am am not sure what to do next:

A) Is it worth doing OTU clustering using one of the three strategies with vsearch? Deblur output should be already dereplicated and filtered for chimeras;

OR

B) Should I just proceed to assigning taxonomy and move to alpha and beta diversity? For what I understand, Q2 developers are trying to move away from OTU clustering using vsearch.

Thank you for your kind attention,
Max

[jwdebelius (Justine Debelius)]

Hi @mstagliamonte,

@mstagliamonte

A) Is it worth doing OTU clustering using one of the three strategies with vsearch?

Nope. Denoising is replacing OTU clustering. I broke down a couple ideas with clustering and denosing here and here a bit ago, so those threads might be helpful.

Deblur output should be already dereplicated and filtered for chimeras;

Deblur's filtering stragedy means that chimeras should get handled there. By the way, chimera slaying is mostly built into DADA2 as well, and is used by default in the QIIME 2 implementation, so if you run DADA2, you don't need to do additional chimera slaying.

@mstagliamonte

For what I understand, Q2 developers are trying to move away from OTU clustering using vsearch.

I think this is the field in general. ASVs give you more specificity at very little cost, which is nice? And so, if its still cheap and mostly feasible to get ASVs, keep most of your data, and combine (most) datasets, why not?

...All of which is a long winded way of saying the option B is probably your best bet.

Best,
Justine

[mstagliamonte]

Hi, Justine,

Thank you for your kind answer. Looking at your links I realized this was not the first time you had to answer this question …

I’m going to follow the tutorial and experiment with deblur and dada2. I definitely need to read and re-read more regarding ASVs.

Thanks again for the explanation,
Max