How to obtain database sequence id instead of taxonomic assignment in taxonomic.qza files

[vval (Vincenzo)]

Dear all,
I was looking for a solution to my problem in the forum, but I didn't find anything similar.
I performed a taxonomic classification on a set of data, and I correctly obtained the taxonomy.qza table. I'd like to know if it's possible to substitute the taxon label with the ID of the best-matching reference sequence.
To be clearer, when I open the taxonomy.tsv file, the output I have is something like:

Feature ID Taxon Consensus
0a0000a00.. k__Bacteria...s__thermophilus. 0.9

what I'd like to obtain is:
Feature ID Taxon Consensus
0a0000a00.. 10000000 0.9

where "10000000" corresponds to the database sequence labeled as "k__Bacteria...s__thermophilus".

Thank you.

[Nicholas_Bokulich]

Welcome to the forum @vval!

@vval

I’d like to know if it’s possible to substitute the taxon label with the ID of the best-matching reference sequence.

This is possible, but only with some simple adjustments — because to clarify the classifiers are designed specifically to find the most likely taxonomic lineage (by consensus of multiple hits) rather than the top hit.

Two things:

1. adjust your reference taxonomy so that the feature ID is repeated in both rows. A quick bash oneliner is one way to do this:

paste taxonomy.tsv taxonomy.tsv | cut -f 1,3 > new-taxonomy.tsv

Import to QIIME 2 (and adjust the header lines if there is a header line)

2. classify using qiime feature-classifier classify-consensus-vsearch --p-maxhits 1

Good luck!

[vval (Vincenzo)]

Thank you, it worked! I'm just curious, why is the option --p-maxhits 1? I tried to classify my reads without the option, but I obtained only "Unassigned".

[Nicholas_Bokulich]

@vval

why is the option --p-maxhits 1 ?

You are forcing the classifier to only take the top hit, so it does not perform any type of consensus classification.

@vval

I tried to classify my reads without the option, but I obtained only “Unassigned”.

Because your "taxonomy" does not consist of a hierarchical taxonomy, it consists only of IDs. The consensus classification finds the common taxonomic lineage of the top N hits that match your query sequence, but this only works if you have a semicolon-delimited hierarchical taxonomy. In your case there is none so if you have more than 1 valid hit the result is "unassigned".