SILVA 138.1 LSU reference HELP

[matthewbur (Matthew Burdick)]

I downloaded “SILVA_138.1_LSURef_NR99_30_06_20_opt.arb.gz” from the arb-silva website to use as the classifier for 18S protozoa data in qiime2 2019.10.

Code:
qiime feature-classifier classify-sklearn
–i-classifier SILVA_138.1_LSURef_NR99_30_06_20_opt.arb
–i-reads rep-seqs.qza
–o-classification taxonomy.qza

Output:
(1/1) Invalid value for “–i-classifier”:
‘SILVA_138.1_LSURef_NR99_30_06_20_opt.arb’ is not a QIIME 2 Artifact (.qza)

Is there some type of way to convert the classifier file from .arb to .gza? Or is there somewhere else I need to download the LSU reference from?

Thanks!

[colinbrislawn (Colin J Brislawn)]

Hello Matthew,

Take a look at the Training Feature Classifiers tutorial

If you can convert the .arb file to .fasta, or download a fasta file directly, you should be good to go!

Colin

[SoilRotifer (Mike Robeson)]

Hi @matthewbur,

@matthewbur

I downloaded “SILVA_138.1_LSURef_NR99_30_06_20_opt.arb.gz” from the arb-silva website to use as the classifier for 18S protozoa data in qiime2 2019.10.

The arb file is not the correct file to use. That is for use within arb. As @colinbrislawn suggests you can download the fasta, and other files directly. For this, I'd highly recommend using RESCRIPt to download the files you need. Note: due to some recent changes, you may need version qiime2-2021.2.

You can follow the approach outlined in this thread:

Weird Taxonomic classification 28S rRNA

I can’t really say, it is largely dependent on the sequence lengths present within the SILVA database. You can also simply filter everything to the same length too via filter-seqs-length. Also, one option is to not filter based on length at all until after you’ve extracted your amplicon region first. That is, you may remove short “full-length” sequences which actually contain your complete amplicon region. We allude to this in the tutorial: As for your next question… The SILVA 138 LSU was n…

-Cheers!
-Mike