Because some people (myself included) were having a hard time training their UNITE classifiers for fungal ITS data due to memory errors, I am making my two classifiers available here for anyone who would like to use them (links created with the help of @gregcaporaso).
UNITE ver 7.2 99% Classifier
UNITE ver 7.2 Dynamic Classifier
Some pertinent information:
There are two classifiers available. One uses a 99% threshold for clustering and the other is ‘dynamic,’ meaning it uses either a 97% or 99% threshold (or somewhere in between) based on which is more accurate for certain lineages of fungi (as determined manually by experts in the field).
The version of UNITE I used is the most current (version 7.2, release date 2017-12-01). I retrieved it from the following site: https://doi.org/10.15156/BIO/587481
I used the most current version of QIIME2 to create the classifiers (2017.12), and I followed a protocol posted previously on GitHub, just updating the source files: https://github.com/gregcaporaso/2017.06.23-q2-fungal-tutorial
The classifiers were created without a “feature-classifier extract-reads” command so they should work well for any primer sets. According to @Nicholas_Bokulich, with fungal ITS data we see a performance decrease when the feature classifier is trained on extracted sequence reads. See this post for more info: https://forum.qiime2.org/t/memoryerror-when-training-unite-ver7-01-12-2017-classifier/2757/2
Please don’t hesitate to contact me if there’s any other information you need!