Gneiss using taxa collapsed table

[Dorothy]

Hi,

I'm trying out gneiss, and encountered another problem which I would like to ask here.

I used DADA2 to process MiSeq paired end data then analyzed with gneiss, however I found that ASVs belong to the same taxonomic level (genus) are present in both numerator and denominator which is very misleading. So I collapsed the table to genus level first, then at balance-taxonomy step encountered another problem because the featuerIDs are no longer matching, I guess there is way go around it.

I'd like to know if it is more reasonable to work on a taxa collapsed table with gneiss to find out which groups of bacteria really differ?

[mortonjt (Jamie Morton)]

Hi @Dorothy, yes working with collapsed taxa is definitely fine, and can also mitigate issues with zeros. However, if you wanted to do that, you’d need to collapse the taxa, and feed that information into the add-pseudocount, ilr-transform , …