Table from dada2 cannot be visualized as a .qzv file


I am trying to find the optimal truncation length in dada2 to obtain as many asv per sample as possible. Hence I want to visualize the table.qza output from dada2 as a table.qzv via the metadata tabulate function, which works just fine.

But when I want to view the table.qzv on qiime-view, it doesn't load until the page expires.
Where's the problem? the rep-seqs.qzv and stats.qzv can be viewed normally, just not the table.

Thanks in advance!

(qiime2-amplicon-2024.2) jonas@Jonass-MBP dissertation % qiime metadata tabulate --m-input-file table-single-reads-110bp.qza --o-visualization table-single-reads-110bp.qzv

Saved Visualization to: table-single-reads-110bp.qzv

Hello Jonas,

Can you post the .qzv file here or send it to me as a DM for testing?

If your browser can open other QZA files okay, the issue must be within this one file.



Thank you for your reply.

I think I found a different way by using qiime feature-table summarize

May I ask a follow-up question?

1.) Am I right to assume that - in order to obtain the biggest amount of asv/ diversity - a dada2-truncation length of 100bp would be ideal?
As far as I understand "number of features" can be translated to "number of asv". And the number of features is highest at the length of 150.

2.) Why does the median frequency per sample drop the longer the truncation length?

Thank you!

Hello Jonas,

Sure, we can have a follow-up discussion here!

This is a trade-off between

  • trimming less to get long reads (good for telling similar microbes apart)
  • trimming more to get higher quality so more reads pass filter (good for measuring smaller changes in microbial composition)

What are you most interested in?

Now, Llt's disambiguate this:

Yes, a 'feature' is an OTU / ASV / sequence variant.
Reads are denoised into ASVs or clustered into OTUs.

Uh, I try to optimize DADA2 to output the greatest number of reads == total frequency of features. Having more ASVs is not necessarily a good thing.

Longer reads will have a higher expected error rate and so more will be filtered out.
Longer reads -> more errors -> fewer reads after filtering -> lower median frequency and lower total frequency of counts per sample

I hope this helps!

Your post helps a lot! Thank you, Colin! I am most interested in small changes in the composition, not into the deepest characerization of Staphylococci :wink:

Having more ASVs is not necessarily a good thing.

Could you please elaborate on this? Why not?

Sure! I'm trying to avoid the 'bigger is better' trap. :thinking:
Instead, I want to focus on 1) explaining the biological context and 2) increasing statistical power. :student:

Having more reads per sample is good because this larger sampling effort captures more diversity within each sample (alpha diversity) and can capture smaller changes between samples (beta diversity.)

Having more ASVs per sample means that the observed_features alpha diversity of that sample is bigger... but is it better?? Maybe! You tell me!