I am working on a support vector machine doing some classification based on microbiomes. I am looking to take the relatedness of OTUs into account via some manner of distance measure, but it needs to be a single value. A distance matrix won't cut it. I have already proven to myself that, for an ingroup (in this case, one of the OTUs in my dataset) you cannot use the distance between that ingroup and every other OTU as a measure because even though two OTUs are similarly dissimilar, they are not guaranteed to have a small dissimilarity (distance).
What if instead of an ingroup, I used an outgroup. A OTU external to my dataset that I know to be on the other side of the root? For discussions sake, lets say a Shark (though I would make my real outgroup more closely related to the microflora). My question is, can I say that two OTUs that have a similar distance from an outgroup (the shark) are also simular to each other? I don't need it to be 100% true all the time, but is this generally true? Any advice or examples would be greatly appreciated. If you have recommendations on how to better ask questions here on Biostars, I am also all ears.
Best!
