Same ASV comes up as a different taxonomy when doing BLAST in comparison to running feature-classifier using SILVA database

[ankurnaqib]

I ran the whole qiime pipeline and I am seeing an issue here, and I am not sure why this is happening. For example, lets take ASVID e4326e04a50bad416b5aa32620852cbd (BLAST result)

But when I run the command
qiime feature-classifier classify-sklearn
--i-classifier /home/anaqib/References/qiime2_2023_5/silva-138-99-nb-classifier.qza
--i-reads dada2-ccs_rep_filtered_v1.qza
--o-classification dada2-ccs_classified_filtered_v1.qza
--verbose
&> DADA2_classifier.log

The output for the same ASV comes as:-

Feature ID	Taxon	Confidence
e4326e04a50bad416b5aa32620852cbd	d__Archaea; p__Aenigmarchaeota; c__Deep_Sea_Euryarchaeotic_Group(DSEG); o__Deep_Sea_Euryarchaeotic_Group(DSEG); f__Deep_Sea_Euryarchaeotic_Group(DSEG); g__Deep_Sea_Euryarchaeotic_Group(DSEG); s__uncultured_archaeon	0.712582

Any clue why this is happening? Any hekp would be really appreciated. Thank you.

[Nicholas_Bokulich]

Hi @ankurnaqib ,
Which region are you targeting/which primers are you using?

Did you train your own classifier or take the pre-trained SILVA classifier?

Strange hits like this to uncultured groups has been reported in the past if very short or noisy sequences are kept in the reference database, or if the incorrect reference database is used.

You are using two different taxonomic classification methods (blastn, naive bayes) with two different reference databases (blast nt, silva) so this is sort of comparing apples and oranges until we know more!