Issue using Silva classifier high number of only Bacteria assigned

[geryale29]

Hello, I have an issue with the taxonomy of silva 132 database in qiime2-amplicon-2023.9.
First, I was analyzing 192 samples of 16s pig's microbiota, when I analysed all the samples at the same time with the classifier, I obtained a high number of D_0__Bacteria;;;;;__, (71.528% higher).
Then I decide to analysed the most 'problematic' samples separated from the rest, suddenly I saw that this exact samples are assigned perfectly when they are separated on another different group:

Here is the command I used to do the taxonomic classification.

qiime feature-classifier classify-sklearn \
  --i-classifier ownSilvaClassifier.qza \
  --i-reads 6.2_rep-seqs-op_ref-85.qza \
  --o-classification 11_taxonomy-sklearn.qza 

[SoilRotifer (Mike Robeson)]

Hi @geryale29,

The likely cause if the fact that many of your sequences are not in the same 5' - 3' orientation as the reference classifier. That is, they are in a mixed orientation, as discussed in quite a few threads.

I refer you to the following threads, as a potential way to get your sequences oriented:

• bullet list / sanity checks
• classifier for forward reads
• preparing mixed orientation reads for dada 2

[SoilRotifer (Mike Robeson)]

@geryale29, I also just realized that you might be using an older version of the SILVA reference database, as suggested by the taxonomy rank:

@geryale29

D_0__Bacteria;;;;;__,

Later versions of QIIME 2 now provide premade SILVA (v138) classifiers with a Greengenes-like taxonomy. These classifiers, and the files used to train them are available via the Data resource page. You can even curate SILVA, and other reference databases, with RESCRIPt.