I’m unsure what data transformations are being calculated prior to the distance matrix generation.
qiime diversity core-metrics-phylogenetic it looks like weighted and unweighted unifrac distances are calculated along with Bray-Curtis and Jaccard and then PCoA matrices of each. It also looks like 'FeatureTable[Frequency]` is used as input.
My understanding is that for both Jaccard and unweighted unifrac, the input feature table would be converted to presence/absence data first. For Bray-Curtis, if a PCoA is being calculated, I’m assuming that the FeatureTable is converted to at least abundance data, as BrayCurtis on count data would violate the triangle inequality (and when Bray Curtis is calculated on abundance data it becomes analogous to the L1 distance?)
qiime diversity beta option, possible distance metrics also include the Aitchison (I’m assuming the clr or ilr is calculated on the FeatureTable first), among several others.
I have been performing a custom transformation (version of the Hellinger) and would like to calculate Jaccard, Bray-Curtis, and weighted and unweighted unifrac on that data. I re-imported the data into qiime and coerced it into the FeatureTable[Frequency] semantic type and the command runs fine, but I’m not convinced it’s not doing additional transformations prior to Bray-Curtis and weighted unifrac (I’m assuming it is converting to presence/absence for Jaccard and unweighed unifrac).
Sorry this is a bit long winded, in summary, for commands in the Diversity plug-in that return various distances using FeatureTable[Frequency) as input, are the data transformations required for each distance wrapped into the function? If so, is there a command (or set of commands) that will calculate the distance with no pre-transformation (input transformed data).