Filtering by pathway abundance

[Alan Hudson]

Hi,

Does anyone have any advice as to whether filtering out low abundance pathways is beneficial post-PICRUSt2? If so, what is the best way of finding an appropriate level of abundance filtering for a specific data set?

I have already filtered low abundance ASVs prior to the PICRUSt analysis. I have seen an arbitrary count of less than 10 used to filter (Bletz et al. 2016, via Phyloseq query #720).

I assume this is done pre-normalization of the pathway count data.

Thanks again for any help,

Alan