Appropriate input data selection for PICRUSt2 analysis

Data resource
1

Hi,

I am conducting a rat experiment in which I compare four different groups and look for differences among them.

I sequenced the fecal microbiome using 16S rRNA, performed ANCOM analysis, and ran PICRUSt2. At the moment, I have run PICRUSt2 only on the bacteria identified by ANCOM as characterizing each group.

My question is whether it is appropriate to run PICRUSt2 only on the bacteria that are unique or significantly associated with each group, or whether I should instead run it on the full bacterial dataset for each group (which would naturally include the group-specific bacteria as well).

Thank you very much

2

Hi @chen_Ifrach,

Typically you would run PICRUSt2 on the total community as its predictions for total community function are fairly accurate but microbe to microbe prediction may be less accurate (especially if looking very specific functions). I would strongly suggest following up any microbe specific predictions with further validations when possible.

As an aside I would also encourage you to use the PICRUSt forum here for questions, the current maintainers are very active at answering questions here.

Cheers,
Jacob Nearing