Silva classifier code is not working

[Lacona]

Hey all,

I tried to solve that problem by myself by searching web and the forum, but I am stuck.
So, I am very sorry, but I have to ask you again:

qiime2-2020.8, miniconda, Mac, 25GB free RAM
Primer: 515f/926r

After training the greengene classifier without any issues (thank you for the gorgeous tutorial!) I tried to do the same with silva, using
Silva 138 SSURef NR99 full-length sequences and Silva 138 SSURef NR99 full-length taxonomy.

First I extracted the reference reads - and it worked nicely:

qiime feature-classifier extract-reads \
--i-sequences silva/silva-138-99-seqs.qza \
--p-f-primer GTGCCAGCMGCCGCGGTAA \
--p-r-primer CCGYCAATTYMTTTRAGTTT \
--o-reads silva/silva_99_ref.qza

silva_99_ref.qza

Then I tried to train the classifiers:

qiime feature-classifier fit-classifier-naive-bayes \
--i-reference-reads silva_99_ref.qza \
--i-reference-taxonomy silva-138-99-tax.qza \
--o-classifier silva_99_class.qza
--qiet

The first try was around 2hrs, the second one around 4hrs and the last one over night around 12hrs. Result of all runs was nothing. I mean, literally nothing happened, not even an error message popped up. I couldn’t stop it running either, not with q, nor command D or command C or kill -9. I had to shut down my laptop to get out.

I sent the code to my PI and he tried to run it for me, but it didn’t work either. In my opinion qiime couldn’t find the files, so we will try again on Monday.
In case it is of any help, I can provide the verbose output.

I am very sorry, that I have to bother you with questions again.
Every kind of help is greatly appreciated!
Andrea

[jwdebelius (Justine Debelius)]

Hi @Lacona,

If the command is hanging, that means it’s running. Depending on your avaliable computation, it can take several hours to days to compute.
If you’ve got multiple threads going, you may need to ctrl+C several times to kill all threads.

BTW, It looks like you might be running the EMP 515-806 primers. There are some pre-trained classifiers available for those primers. A casual user can usually get away with the pre-trained classifiers. (Heck, I rarely train my own classifier these days, it’s not worth the time/effort.)

Best,
Justine

[Lacona]

Thank you for your answer.

It runs honestly for more than 12hrs?!
Ok, next time I will try command-C more often.
Unfortunately my primers are 515 and 926, therefore I had to train the classifier by myself.

[jwdebelius (Justine Debelius)]

I don’t know about your computational resources or your database, but it’s possible. In your shoes, I think would just use a full length classifier. It gets you on your way faster.

Best,
Justine

[Lacona]

Thank you!
I will try the silva-full-length-sequences classifier then.