Using QIIME1, I used an open-reference OTU picking method, and for downstream PICRUSt analysis, was able to filter out the de novo sequences leaving just the OTUs with matching IDs present in greengenes database. With QIIME2, first I can’t find any information on what OTU picking method is being used. Second, the OTU IDs are named completely different containing random numbers and letters. Can someone please help me to understand 1) what OTU picking method is being used by QIIME2, and 2) how can I filter out de novo OTUs and get OTU IDs match with the GG database for PICRUSt analysis.
At the moment in QIIME 2, we’re not doing any OTU picking, but rather working with sequence variants (i.e., 100% OTUs). This pre-print provides a good description of the motivations for this. Sorry that you weren’t able to get this info from the docs - I’ve created an issue to clarify that.
For the time-being the feature tables that are generated with QIIME 2 are not compatible with PICRUST - you’ll still need to follow their instructions for picking closed-reference OTUs (or using your procedure of open-reference OTUs filtered to only the GG OTUs) to run PICRUSt. We’ve been talking with Morgan Langille, the lead developer of PICRUSt, and he is interested in developing a QIIME 2 plugin for PICRUSt, so you can expect that in the future.
Sorry for the inconvenience!
An off-topic reply has been split into a new topic: When will a PICRUSt plugin be developed?
Please keep replies on-topic in the future.
Quick followup for anyone stumbling on this topic:
As of the 2017.9 release, qiime2 supports de novo and closed-reference OTU picking via the
q2-vsearch plugin. There’s a community tutorial with details:
With this new OTU picking support, it is now possible to perform closed-reference OTU picking with
q2-vsearch and then export the feature table for PICRUSt analyses. See this forum topic for details: