Combining 16S and ITS data together?

Hello everyone. I was wondering if there was a module or a method to combine both the 16S and ITS data together in order to perform the standard analyses? I have been doing it separately so far (different analyses for the 16S and ITS data) but I am interested to know if my samples might differ with the combination of the two data sets. Does this make sense to do this and if so is it possible within the qiime framework?


Hi @vpandyar,
Sure, you can do this: use qiime feature-table merge to merge those feature tables together. That will work as long as the feature tables have the same sample IDs.

Whether this makes sense is another question. I will often merge 16S and ITS data (or other marker gene data) for analysis with q2-sample-classifier. This could also arguably make sense for making PCoA plots, though you may need to normalize the data before merging. For alpha diversity and other analyses I doubt it would be valuable to use merged data.

Thank you for the prompt reply! For my own edification, how would I go about normalizing the data prior to combining it? Also why would the combined data not be useful for alpha diversity analysis? Presumably it is okay for beta analysis given that you say PCoA plots are possible?

you could rarefy at an even sequence depth

Imagine you have two samples, one with 99 fungi and 1 bacterial ASV, the other with 99 bacteria and 1 fungal ASV. After combining, you would observe 100 ASVs in each, hiding an important difference between these samples. Also, merging your 16S and ITS data would prevent you from using phylogenetic methods (both alpha and beta diversity)

I said it would be possible, but I don't want to endorse it as being okay. In general, keeping the data separate is likely to reveal more information for alpha and beta diversity analyses.

Thanks for the response. This was actually really informative. The reason I am asking these questions is that I could not detect differences in diversity either with the 16S or ITS data I have but was wondering if by using the combined data I could possibly detect differences in samples. Sounds like combining the data in this way is not useful to do this type of analysis. Are there any other useful measures or methods I could use to detect differences in the sum total of all my data or is separate analysis always the best?

Thanks again!

If neither shows differences, then combining them will probably not reveal anything new via diversity analyses. However, it would be worth combining the feature tables and using q2-sample-classifier to see if you can differentiate groups.

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.