Converting ASV from DADA2 to OTU

Hi,
I am relatively new to Qiime so I apologise for any simplicity in the question.
Is it possible to convert the output of DADA2 (ASV) table into an OTU table. From what I read there seems to be conflicting opinions. I think I can use the v2-search?
Or am I better off just removing the DADA2 pipeline and using something else entirely ?

Thanks for any help in advance :slight_smile:

Hi @gdunne

Yes, you can convert your DADA2 output in OTU table.

First, you have to export your FeatureTable[Frequency] in BIOM file:

qiime tools export --input-path output_DADA/your_FeatureTable[Frequency].qza --output-path data_exported/

Second, you have to convert your BIOM file in ASV table:

biom convert -i data_exported/feature-table.biom -o data_exported/ASV.tsv --to-tsv

Hi @gdunne,

You can use the command suggested by @Matteo_Scanu to export your table into biom format for other tools, but if the features in the initial table are ASVs, you will always have ASVs.

Just a quick summary, both ASVs and OTUs can be represented within a feature table produced by qiime2, but these are different way to denoise your sequences. ASVs are obtained after an error correction step, OTUs after a clustering of your sequences using a selected similarity threshold (could be either a de-novo clustering or onto a reference they are still called OTUs and may be bit confusing).

So, you can use your dada2 output sequences to further cluster them by q2-vsearch (as you suggest), so that you obtain OTUs.
The question is if you do really need to work with OTUs. Anything you could do with OTUs you can also do with ASVs, but OTUs are dataset-dependents (if you add more sequences to data resulting OTUs may change), ASVs are dataset independent (if two independent datasets contain the same sequences they should produce the same ASVs *), so I usually work with OTUs if I got a specific need (that is very rarely these days).

  • Caveat on ASVs, on dada2 the denoising step depend on a random chosen subset of sequences to asses the errors, hence if the two subset, one for each datasets, will be very different in quality or type of sequences it may happen that the resulting ASVs will show some unexpected differences.

Does this help you in your choice?
Cheers
Luca

1 Like

Hi @Matteo_Scanu
Thank you so much for this, this is very helpful! :slight_smile:

Hi Luca,

This is extremely helpful thank you so much. I now think I've got a good grasp on the distinction between ASV and OTU!
The current debate is ASV vs OTU, however, I did think that since I wouldn't be adding any more sequences to my dataset that OTU might be better to use.

Thanks so much again! :slightly_smiling_face:

Hi @gdunne
personal not requested suggestion, use ASVs anyway, so your sequences have been denoised. Also in future you never know what is going to happen, and you will be bale to compare this with other dataset without the need to reprocess them. Plus, is a newer approach and reviewer may complain on the use of OTUs (which is a valid approach, nothing wrong on using it but just to avoid the discussion that may be a waste of time)
Luca

1 Like

Hi @llenzi,

Thank you so much for the advice!
All the best :smiley: