feature classifier and pre-trained classifier

Hi there,

  1. I’m not sure in what case I can use pre-trained classifier(gg-13-8-99-515-806-nb-classifier.qza)?
  2. If my primers are the V4 primer on EMP 16s protocols(http://www.earthmicrobiome.org/protocols-and-standards/16s/), can I use pre-trained classifier?
  3. qiime feature-classifier extract-reads command
    Is it all rigtht if I don’t set any optional pamameters? These parameters: --p-trunc-len, --p-trim-left, --p-min-length, --p-max-length, I have no idea what value fit best. For example, my sequences have been trimmed to 150bp in previous deblur step, now how can I set these parameters?

Thank you very much for your kindly help in advance!


Hi @maque4004,

If your target amplicons match or are internal to the sequences used to train the classifier. So if you use any 16S amplicon, you can use the full-length classifiers. Or…

Yes, you can use either of the classifiers trained on amplicons targeted with these primers, e.g., gg-13-8-99-515-806-nb-classifier.qza

Yes, see the notes in the “training a feature classifier” tutorial.

You could set trunc-len to 150 and it may slightly boost classifier accuracy — but the boost would only be ever so slight so if you are using the 515f-806r primers (EMP) you should just use the pre-trained classifiers and save some time!

Thanks a lot! @Nicholas_Bokulich
My primers are V4 515f-806r primers(EMP), I can use both Greengenes 13_8 99% OTUs from 515F/806R region of sequences and Greengenes 13_8 99% OTUs full-length sequences pre-trained classifier, right? Does the 515F/806R pre-trained classifier classify better than full length pre-trained classifier? What’s the difference of using these two pre-trained classifier?


very slightly better yes

see the original publication for q2-feature-classifier to learn more.