Number of assigned taxa


Using qiime1 I got my de novo otus and obtained 453 taxonomic assignments at the genus level using greengenes database. Now I am working with the same files in qiime2. I imported rep seqs from qiime1 in order rep-seqs.qza files and work with the same 97_otus and 97_otus_taxonomy fasta files. I trained the Naive Bayes classifier and run the taxonomy analysis. This time I got 1485 taxa at the genus level. Is that expected? Any ideas why such difference?

Hello Oscar,


A lot has changed between Qiime 1 and 2, including introduction of new taxonomy assignment methods and replacement of OTU clustering algorithms with denoising methods like DADA2 that produce ASVs (instead of OTUs). All of these things will create results that are dramatically different (and hopefully better!).

Let me know if you have any questions about this process.


P.S. It’s worth noting that biological conclusion should remain similar; the same major taxa should be found by both methods and similar clustering of samples should be observed.

Short answer: yes, this is not surprising

What is the total number of query sequences?

These counts — 453 and 1485 — are the counts of unique taxonomic labels that you are observing, correct?

As @colinbrislawn indicated, QIIME2 has had a number of improvements since QIIME1 (though it sounds like the only difference in your data is the taxonomy assignment methods, not denoising). The classifiers in QIIME2 are one of these improvements, and we have optimized these methods to be a bit higher precision than the defaults in QIIME1 (though those classifiers do still perform quite well).

My guess is that you are seeing more unique taxonomy strings with QIIME2 because the naive-bayes classifier is successfully classifying your sequences at deeper taxonomic levels (e.g., species instead of genus/family). Without knowing the true composition of the samples that you are analyzing, it is impossible to say whether this is actually “better” or “worse”. But it should at least be satisfying :slightly_smiling_face:

Does that seem to fit with what you are seeing? Or do you think there is something more sinister afoot?

Yes @Nicholas_Bokulich I did the denosiing steps in qiime1 so only the taxonomy analysis was done in qiime2 and as @colinbrislawn says, the biological conclusion is the same as with qiime1. The taxonomic structure remains the same only that now I got much more labels. “Unassigned” also got reduced from ~2% to ~0.01%. I was a bit suspicious but now I know those results are great! Many thanks!


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