How to generate confidence value from consensus-blast?

[starlight]

Is confidence supported in qiime feature-classifier classify-consensus-blast as it is in qiime feature-classifier classify-sklearn? We are working in qiime2-2023.5 with a custom 12s classifier. Blast seems to output better classification but no confidence scores while sklearn outputs confidence but poorer classification.

[cherman2 (Chloe Herman)]

Hi @starlight!
Welcome to the forum!
I don't believe that confidence is supported for qiime feature-classifier classify-consensus-blast. You can mess around with many parameters in qiime feature-classifier classify-consensus-blast to guarantee more strict guidelines for taxonomic assignment.

here are some parameters I would look into:
--p-maxaccepts
--p-perc-identity
--p-query-cov
--p-evalue
--p-min-consensus

Hope that helps!

[Nicholas_Bokulich]

Hi @starlight ,
To add to @cherman2 's great answer: no, indeed the classify-consensus-blast action does not output a confidence score, as the classifier operates in a very different way from the classify-sklearn classifiers.

@starlight

We are working in qiime2-2023.5 with a custom 12s classifier.

For both classifiers you should play around with the different parameters to see if you can improve the results. The default parameters for both were set based on performance with 16S and ITS data, so for a 12S classifier you will need to re-tune some parameters to optimize performance, I would not expect it to necessarily work well with the default parameters.

Good luck!