Goin from OTU clustering to feature classifier


I'm encountering difficulty when trying to feature-classify my de-Novo OTUs. My Goal is to eventually have frequency data per sample for my different OTUs while including the taxonomic ID generated from feature classifier. I would do this by exporting those two separate artifacts and then merging them together into one excel file.

The problem is this: My feature IDs generated for my OTUs do not match the feature IDs generated after running feature-classifier, so I have no way to know which OTUs (and therefore the frequency data) align to what taxonomy.

I ran a feature classifier several weeks ago on the feature table generated from closed and open clustering and I didn't have this problem, but the OTU centroid/accession species generated seemed to be completely novel- but thats another issue.

Can anyone tell me why my feature IDs are changing when I run the feature classifier command? Is it something to do with the input reference taxonomy I have to use to run the command?


Hi @KWyssmann ,

The feature IDs change when performing any type of OTU clustering, either to match the reference IDs (closed reference clustering) or to assign arbitrary IDs (de novo or open-reference clustering).

However, the IDs should not change when performing taxonomic classification. Could you please provide the QZAs of the data that you are trying to link? (QZVs would be helpful also, e.g., run qiime feature-table summarize on the tables and qiime metadata tabulate with the taxonomy as input)


1 Like

I looked at my DADA2 output and the results of my de-novo clustering and see that those outputs share the same feature IDs (the de-novo results show less features than the DADA2 output which would be expected after clustering). However, when I run the feature-classifier-consensus-blast plugin all the feature IDs change.

Here's what I ran:
qiime feature-classifier-classifier classify-consensus-blast
--i-query CoxDeNovoOTUseqs.qza \
##my DeNovo clustering results
--i-reference-reads NCBIBoldCoxseqs.qza \
--i-reference-taxonomy NCBIBoldCoxTaxonomy.qza
##these qza's were created by running the rescript command to pull NCBI reference data using keywords Cytochrome c Oxidase subunit/COX1/CO1/ marine etc
--p-maxaccepts 1
--p-perc-identity 0.90
--o-classification CoxDeNovoOTUTaxonomy.qza
--o-search-results CoxDeNovoOTUtaxclasssearchrsults.qza

Nevermind! It was a simple mistake and I figured it out! Thanks!

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