Training a new classifier: primer and reference sequence

[Charlie]

Hello!

Now I am trying to train my classifier. I downloaded the 13_8 database and I want to use 97%, but I'm not sure which fasta file should be input when using these commands:
qiime tools import
--type 'FeatureData[Sequence]'
--input-path 85_otus.fasta
--output-path 85_otus.qza

1. /gg_13_8_otus/rep_set/97_otus.fasta

1111883
GCTGGCGGCGTGCCTAACACATGTAAGTCGAACGGGACTGGGGGCAAC

2. /gg_13_8_otus/rep_set_aligned/97_otus.fasta

1109237
----------ATTCCGGTT-GA----------

The second question is about this command:
qiime feature-classifier extract-reads
--i-sequences 85_otus.qza
--p-f-primer GTGCCAGCMGCCGCGGTAA
--p-r-primer GGACTACHVGGGTWTCTAAT
--p-trunc-len 120
--o-reads ref-seqs.qza

I asked the person who constructed the sequencing library, but he said they did not use any primers, and the primer sequence seems to be not necessary in Qiime1. Can I train the classifier without primer sequences?

Thank you very much!

[Charlie]

Hello everyone:
I asked the sequencing library constructor again, and he told me the primer sequences!! Hope it works. Moreover, I guess I have to use the 97_otus.fasta in /rep_set directory because the other one doesn’t work.

[Nicholas_Bokulich]

@Charlie

I guess I have to use the 97_otus.fasta in /rep_set directory because the other one doesn’t work

yes, you need to use the unaligned sequences.

@Charlie

–i-sequences 85_otus.qza

Don't use the 85_otus, there is not enough specificity. That file is only used in the tutorials for the sake of speed. Use the 99_otus instead.

@Charlie

the primer sequence seems to be not necessary in Qiime1

The primer sequences are not necessary here either, but it does improve classification slightly as noted in the tutorial that you are following.

QIIME2 is a whole different animal from QIIME1... so just because something was done a certain way in QIIME1 does not mean that it is the right way to do things. Many things have changed, and all for the better

Good luck!

[Charlie]

Thanks for your detailed reply! As you mentioned primer sequences are not necessary in Qiime2, how can I train classifier without primer?
I guess:
qiime tools import
–type ‘FeatureData[Sequence]’
–input-path 97_otus.fasta
–output-path 97_otus.qza

qiime feature-classifier fit-classifier-naive-bayes
–i-reference-reads 97_otus.qza
–i-reference-taxonomy ref-taxonomy.qza
–o-classifier classifier.qza

Is this right?

Another question is that I used my primers to train the classifier, but it output:
Plugin error from feature-classifier:
No matches found
Debug info has been saved to /tmp/qiime2-q2cli-err-vcz7amvb.log

my primer is Illumina 16s Amplicon PCR Primer
–p-f-primer TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCAG
–p-r-primer GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC \

Are they too long so no matches found? If so, how long should I trim it.
Thank you!

[Nicholas_Bokulich]

@Charlie

Is this right?

yes

@Charlie

Are they too long so no matches found? If so, how long should I trim it.

Yes, those are adapter + primer sequences. See here.