UNITE classifier

bsen2018 · February 7, 2018, 10:08am

Hello!
Because of the MemoryError issue (as I am working on a Windows machine with only 6 gb RAM) while training feature classifier for the latest UNITE database (version 7), I am unable to generate the UNITE classifier.qza. I would like to know if someone has prepared one and if it is possible to have the UNITE classifier.qza like Silva, Greengenes from QIIME2 data resources, so we can download it for further taxonomy analysis in QIIME2?
Thanks!

colinbrislawn · February 7, 2018, 7:15pm

Great question! I would love to know how qiime 2 supports pretrained / indexed databases.

Colin

gregcaporaso · February 9, 2018, 7:54pm

Hi @bsen2018 and @colinbrislawn,
We would definitely like to have pre-trained classifiers for UNITE available to QIIME 2 users. I just sent an email to the developer of UNITE to see if they would prefer to create and host those on their end in the QIIME release of the database, or if they would prefer if we train and host those on our end using their latest QIIME release. I’ll follow up when I hear back on that.

Sydney_Morgan · February 10, 2018, 4:25am

Hi @bsen2018, I’ve prepared two versions of the UNITE classifier from the latest UNITE database. I had the same MemoryError as you but was able to gain access to a more powerful computer to run the classifier commands. I have a 99% classifier (unite-ver7-99-classifier-01.12.2017.qza) and a dynamic classifier (unite-ver7-dynamic-classifier-01.12.2017.qza). I can send them to you if you would like?

Jaroslaw_Grzadziel · February 10, 2018, 8:26am

I would be also interested in dynamic version

bsen2018 · February 10, 2018, 9:12am

Thanks for taking an initiative on this issue.

bsen2018 · February 10, 2018, 9:16am

Thanks@Sydney_Morgan! I would certainly like to have the two versions of the UNITE classifier that you have prepared, before they are made available in the QIIME2 Data resources. Please let me know how you would like to send it across.

gregcaporaso · February 12, 2018, 7:53pm

I spoke with the UNITE developer, and we are going to discuss how to handle these releases in the future, but their main programmer (who would handle this) is currently out on leave for a couple of months.

In the meantime, @Sydney_Morgan, if you are interested in sharing your classifiers, we could host them for you. This would be really helpful for us, and for a lot of QIIME 2 users. If you wanted to do that, maybe you and I can connect by direct message so you can get me the files, I could post them somewhere and get you links, and then you could post a new forum topic where you describe how the classifiers were generated (e.g., version of UNITE, version of QIIME 2, …) and share the links. I would also be happy to create that post if you just shared the files with me, but if you create the post it’ll be more clear that you should get the credit for creating the classifiers.

Jaroslaw_Grzadziel · February 13, 2018, 12:34pm

It’s a great idea! I have tested classifier prepared by @Sydney_Morgan and everything works perfect in the latest version of Qiime2

bsen2018 · February 13, 2018, 1:10pm

Many thanks to @Sydney_Morgan for sharing the classifier that she prepared. It worked well for me. Hope others too can avail it from QIIME2 data resources soon!

colinbrislawn · February 13, 2018, 11:47pm

For reference, see:
https://forum.qiime2.org/t/unite-ver-7-2-2017-12-01-classifiers-for-qiime2-ver-2017-12-available-here/?source_topic_id=2919

system · March 17, 2018, 6:52am

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.