About PICRUST2 16s rRNA gene database for training representative sequences

[5 skit343]

Hello,

I am writing this post to clarify some points while using the PICRUSt2 plugin in QIIME2. I am currently analyzing 16S rRNA gene data using a rep-seq trained naive Bayes classifier model based on the Greengenes2 database.

While reviewing the PICRUSt2 code, I noticed that the input files include the feature table and rep-seq files, but not the taxonomy.qza file used to train the rep-seq.

Therefore, I assume there must be a classifier within PICRUSt2 that has already been trained on some database. Could you please let me know which database it uses? Additionally, if the data is not based on Greengenes2, is there a way to use my taxonomy.qza file, which is trained on Greengenes2, with PICRUSt2?

Thanks.

[Robyn Wright]

Hi there,

PICRUSt2 actually doesn't really classify sequences. As you can see in the main flowchart, the sequences are placed within a phylogenetic tree. Information about the placement and the genomic content of the closest branches are then used to predict the functions present within your sequences. It is therefore not necessary to provide taxonomy information to PICRUSt2. The genomes that PICRUSt2 works with came from the JGI database. I am currently working on building a new reference database that would contain all genomes from the GTDB database, but will update everyone when that is ready for testing (hopefully soon!).

The only way currently to use the Greengenes2 database would be for you to annotate all of the genomes contained within that database and replace the PICRUSt2 reference data with these files when running PICRUSt2. I don't think that it's possible to specify custom reference files with the QIIME2 plugin, and I think that you'd need to do this with the standalone PICRUSt2 version.

Robyn