DADA2 Error rates could not be estimated

Hi all,

I encountered a DADA2 error where:
3) Denoise remaining samples .................................................Error rates could not be estimated.
but the qiime dada2 denoise-paired command completed successfully, and there is the table.qza and rep-seqs.qza. So I guess these two files were not error-corrected? Then I shouldn’t use these, right? Why was the error not caught, does this mean it is not necessary to run dada2?

In the dada2 github page I found a relevant thread but the cause of the error seems different. @benjjneb was there. The solution was to upgrade R packages but I can’t find e.g. biostrings in the R provided by QIIME2’s miniconda distribution.

Curiously when I ran a subset of the same data I didn’t get this problem.

Thanks a lot.


Check this thread, you’ll probably find the solution.

Previously, I had similar problems that error rates didn’t converge during DADA2-denosing and pair-ended reads didn’t merge. After removing the primer sequence in the reads, the problems were fixed.

That message means that the post-hoc re-estimation of the error rates failed in one of your samples. When you ran a subset of the data, it probably did not include the offending sample, thus no message. My best guess is that there is one particularly small sample that causes the loess fitting to break, but that’s just a guess as I’ve never seen this message crop up in this way before.

But, good news!, it doesn’t matter at all. The post-hoc error rates re-estimated in step 3) Denoising are not used for anything by the QIIME2 plugin. The inference of sequence variants is totally unaffected, as it is based on the error rates estimated in the prior step 2), not the per-sample re-estimates.

In short, ignore that message, it has no impact on your output.


This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.