Are the Shannon Entropy and similar metrics calculated by powerTCR robust against sequencing depth?

[johan-gson]

Hi @hillarykoch,

I'm trying to compare the TCR repertoires from a bunch of samples, and the number of cells per sample vary from ~350 to ~4000. I tested powerTCR on these samples, and to my understanding this works, I understand that the distances between samples are not biased by the number of cells per sample. So, I can see that samples group the way I expected, but I also want to do a statistical test of a metric that represents something that people know what it is, such as Shannon Entropy. What I don't get is if that is sensitive to sequencing depth, or if that somehow is compensated for in your model? Could you give some advice on a suitable metric?