Comparison results from two different trained classifiers

[mefistofele82]

Hi all,

this is a more general question.
I have a 16S dataset already analyzed with QIIME2 using a trained SILVA classifier in 2018.
Now, I have a new dataset that I’m analyzing with a new trained SILVA classifier.
Is it correct comparing the results from the two classifier? or I should reanalyze the old dataset with the same classifier of the new one?
Thanks in advance
Riccardo

[colinbrislawn (Colin J Brislawn)]

Good evening Riccardo,

I think the best option would be to repeat classification so you are using the same database release and classifier for all of your data.

The results should be similar between a newer and older SILVA, but it’s always best to be consistent. (And you know reviewer 3 is going to mention if you use different classifiers !)

Colin

[mefistofele82]

Yes, this is my concern.
Moreover, I think the denoising step with DADA2 could affect results
Thanks for your reply