Use Classifier for trimmed v regions

Dear all,

I have some datasets from different v regions e.g v3-v4, v4 for analysis. The protocol I'm following is that to trim the the v3_region from v3_v4 region so to retain onlny the v4 region. The way i do that, is the use o cutadapt trim-paired using the 515f primer and 806r primer

So here is question, on classifier step after I trim off the primers can i use a pretraining classifier with out primers?
or build my own classifier with out primers? For example using the GenBank with RESCRIPt. Or other database.

And something more i read from many post that, combine datasets from different v region suggests the fragment-insertion step. can anyone explain me after that step which database we use? because we have multi v regions.

Τhanks for the help!.

Hi @iordanis ,

Yes, definitely. Training a region-specific classifier is even shown in the RESCRIPt tutorial on the forum, e.g., here (shown with SILVA, but you can do the same with an NCBI classifier following get-data-ncbi):

It sounds like you may be looking for q2-sidle, depending on the V regions that you have. But if you have a bunch of sequences from different V regions mixed in a single file you could technically also use a full-length 16S classifier to taxonomically classify all of these.

Good luck!

1 Like

Good evening, @Nicholas_Bokulich
thanks for the response!
can i
human-stool.qza (204.9 KB)
also use your taxonomist that you have prepared for human_stool?
My data is for human gut.

And for q2-slide all my datasets i import Separately, after the donise step i will merged them all, still i can use the q2-slide?

Yes, the pre-trained classifiers on the QIIME 2 website are for general use, they are not specific to one environment. So these should work for human stool.

Good luck!

1 Like

@Nicholas_Bokulich

wonderful thanks for the help!

Hi @iordanis,

I would be careful with Silde and multiple regions; I've had issues with diversity being inflated because the algorithm couldn't solve the multiple regions. I think the classifier approach may be better.

Best,
Justine

1 Like

Good morning @jwdebelius ,
Thank you for your advise.
I will try only the classifiers!

1 Like

@Nicholas_Bokulich good evening,

I’m trying to train the gg_138 V4 reference database with the human-stool.qza as the class-weight within qiime feature-classifier fit-classifier-naive-bayes.

I download the gg_138 i create the ref-seqs-v4.qza ref-taxonomy.qza
and now i try to Retrain the Classifier with bespoke waights human-stool.qza from

The command that i use is,
qiime feature-classifier fit-classifier-naive-bayes
--i-reference-reads ref-seqs-v4.qza
--i-reference-taxonomy ref-taxonomy.qza
--i-class-weight human-stool.qza
--o-classifier human-stool-classifier.qza

When i run this command comes out this messeg:

Plugin error from feature-classifier:
Number of priors must match number of classes.
Debug info has been saved to /tmp/qiime2-q2cli-err-yaqv434y.log

any suggestions?

Hi @iordanis ,

This means that the weights and database do not match. The readytowear repository contains the sequences and taxonomy that correspond to those weights, so you will need to train a classifier using those files if you want to use the readytowear weights.

Good luck!

3 Likes

@Nicholas_Bokulich good morning,

Thank you for your advice!

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.