Hi
I am using qiime2-2020.6. I have used Silva 138 99% OTUs full-length sequences from the qiime2 page (https://data.qiime2.org/2020.8/common/silva-138-99-nb-classifier.qza) to classify 16s sequences for mock bacterial communities from Zymoresearch.
However, that classifier is not assigning “Salmonella” at genera level which is expected to be in the community. https://files.zymoresearch.com/datasheets/ds1706_zymobiomics_microbial_community_standards_data_sheet.pdf
I have also tried all possible versions of self-trained Silva 138 V3-V4 classifiers using three different dereplicating mode “uniq” “majority” and “super” in RESCRIPt I blasted Not_assigned ASVs to NCBI nucleotide database then I found majority of “Not_assigned” have hits with salmonella genus.
In contrast, when I classify the same sequences with Silva 132 V3-V4 classifier, it gives me “Salmonella” at genera level.
I also classify some other sequences from other studies too using /silva-138-99-nb-classifier.qza, however, in every case I am missing salmonella.,
The overall classification with Silva 138 classifier is better than Silva 132; but why I am missing salmonella, which is a bit weird for me.
Can you please give me a clue regarding this issue?
This is indeed strange. There are quite a few Salmonella sp. within the SILVA 138 reference database. However, there was a recent correction and update to this database, SILVA 138.1, which may fix a few issues. You can use the approach outlined here to make your own SILVA 138.1 classifier. You can see how to do this if you click the drop-down menu “The gritty details” under the section “Getting SILVA data: Hard Mode”.
Or, simply run rescript get-silva-data with default settings. Though… given what I will discuss below, I think the issue is a problem with genus rank annotations.
I observed that there are some taxonomic inconsistencies at the genus rank for Salmonella. For example:
As you can see the s__Salmonella_enterica taxa have different genus labels, i.e. g__Escherichia-Shigella, and g__Salmonella… and some others (not shown). I think what is happening is that the conflict with the genus labels forces the classifier to only report the LCA f__Enterobacteriaceae. That is, the problem is the inconsistent annotation of the ranks (again see the description under “caveat emptor!” I mentioned above). You may have to manually edit these labels so that they are all consistent, and re-import the modified taxonomy file and retrain the classifier.
Note, other than providing options to 1) use the organism name as the species label, and 2) propagate taxonomy to fill in empty ranks (available in the latest version RESCRIPt on github ), we do not curate the SILVA taxonomy.
Hopefully, we’ll be able to add tools to aid in curating these issues in the future. Please keep us posted.