I noticed that when I run my data to perform taxonomy analysis (as in the 'moving pictures' tutorial), I'm running the analyses on my feature table that I obtained after DADA2 pipeline but before I filtered the sequences based on sampling depth. So, although my diversity analyses are based on only the sequences that have passed the sampling depth value that I chose, the taxonomy analyses are based on all samples and all sequences. Is this correct?
Is there an advantage of running the taxonomy analyses on only those samples/ sequences that were retained after choosing a sampling depth? And if so, is there a way of doing that?
Let me start saying that alpha and beta analyses are not connected in anyway to the taxonomies given to each feature/sequence. In other words, you don't need to assign taxonomy to perform any of those other analyses as they are completely independent.
With that in mind, we normally suggest classifying/assign-taxonomy to all the features as you can do this in a single go and then reuse that information with your rarefied or non-rarefied tables. I guess the next question is which one should you use (rarefied or non-rarefied). Well, that depends on the tool, method or algorithm you are planning to apply. For example, for differential abundance is recommended to not rarefy but only remove samples with really low sequence counts.