are there BLAST classifier and VSEARCH-sklearn parallel versions?

Hello,

I realized there aren't any indications of a parallelized version of the commands

qiime feature-classifier classify-consensus-blast
and
qiime feature-classifier classify-hybrid-vsearch-sklearn
is there any parameter that perhaps is not displayed or even a version of qiime where these commands are implemented with the possibility of using multiple threads?

Hello,

please, run the commands with --help flag. Developers of the plugin provided relevant information there.

Thank you
V

Yes I understand, but the thing is in the code at the feature classifier repository in github, in the file _blast.py it was mentioned that the multithreading with the blast classifier was disabled, what I'd like to know is if there's a way I can access a version of Qiime2 where the parallel version of qiime feature-classifier classify-consensus-blast is available

Hi @caemuller ,

This file does not say that it is disabled, the note in there says that:

num_thread is not exposable

So currently there is no QIIME 2 version with multithreading enabled (in some very old versions there is a thread parameter, but it is ignored by blast).

It would be possible to enable multithreading but only with a significant amount of work. We are always open for contributions. :wink:

I recommend using classify-sklearn or classify-consensus-vsearch. Both are more accurate, parallelizable, better.

The hybrid vsearch method is parallelizable, see the --p-threads parameter. But this method is sort of experimental and tricky to use, as the vsearch part looks for exact end-to-end matches.

hi @Nicholas_Bokulich ,

I really appreciate your quick reply, it's pretty unfortunate and maybe I can check the code of the Blast classifier and try to work on something although as I'm not entirely working on bioinformatics I'm not sure how relevant it would be and how useful it'd be to the community but that's all I had in mind.

Thank you for your help!

Hi @caemuller

Yes I agree. This is why we have never enabled this... I think you are maybe the 2nd person in 6 years to ask about this. We will probably get around to adding this feature, but it is a low priority for this reason.