# How do I measure genetic distance?

I'm not sure how to measure genetic distance. There seems to be many different equations out there, and all the ones I found are rather old.

In my specific problem I want to see if the twist in collagen differs if the species are different. (I have a program to predict collagen twist according to amino acid sequence)

I therefore want an objective numerical value to base my comparaison on. "Nei's standard genetic distance" seems good, but it is very old so I was wondering if there was a modern alternative. Is there an R package or something like that to calculate it automatically?

There are a few answers here, it is a little unclear which is closest to what you want.

[edited to be more complete and direct]

1) Nei's distance is mostly for comparing populations of the same species, so it should be fine if that is what you are doing. I am guessing that based on your post, you are more interested in cross-species comparisons. Of course, the logic is the same, it's just the math is different. For among-population divergence, Fst (fixation index) is also commonly used. But if you are using a sequence alignment or some such data to compute genetic distances, there are lots and lots of metrics. Some of them are very simple- for instance you could use Hamming distance and just count the number of residue differences between each sequence.

As for computing these metrics, there is a webserver called GenePop online that will compute Fst for you if you plug in a dataset. I don't know of R packages that compute these things, I would guess just by looking at wikipedia that coding functions to do this would be relatively trivial (if you are comfortable with r).

2) If you are truly looking among species, a somewhat more appropriate approach might be to model your species in a phylogeny, rather than to use distance methods (though the two are attempting to solve sort of the same problem). There are lots of easy-to-use online programs for this, for instance: FastML. You can just copy-and-paste a seq alignment in there. After you have the tree in for instance newick format, there are good tools for visualizing trees both online at iTOL, and through R with APE.

APE also has good utilities for modeling phylogenetic processes, so you could in principle directly model changes in protein folds through evolutionary time (assuming you have a meaningful summary of protein folding, either continuous or discrete).

I don't have enough reputation to link to more of the resources I mentioned, sorry.

• I've already worked with SeaView for making phylogenetic trees but I am looking for a good numerical value I could use to do some data analysis (and very basic machine learning) Since time of divergence and simply counting different bases seemed a bit basic I am looking for something more accurate... I'm using amino acid sequences of collagen taken from PDB. Thank you, your answer gives me a few pointers. – cachemoi Mar 7 '16 at 22:58
• ah- i see. my naive guess is that all of the distance metrics will tell you similar things. nonetheless, since you are working with different species and want a more sophisticated distance metric, you might try something like protdist from [phylip] (evolution.genetics.washington.edu/phylip/doc/protdist.html), which has various substitution models available. Again, I would recommend trying to model the problem phylogenetically, which is preferred in the field. Distances can be tricky because observations are non-independent. But probably most methods will give the same. Good luck! – Maximilian Press Mar 8 '16 at 5:54

I think my approach would be to make a phylogenetic tree and color the tips by collagen twist. There are several R packages to help with this, including APE, ADE4, and poppr. To make a phylogenetic tree, you first calculate a distance. You mentioned Nei's, and that's actually still used pretty commmonly. I know you're not doing population genetics, but check out this tutorial: https://grunwaldlab.github.io/Population_Genetics_in_R/Pop_Structure.html. They calculate Gst (mentioned in the other comment), talk about a few different distant metrics (including Nei's), make phylogenetic trees, and it's all in R. They even color the tips by population - in you're case, your populations would be types of collagen twist with your samples named by species (Lizard-1, Lizard-2, Chimp-1, Chimp-2, etc.). You also could color the tips by species and name your samples by twist type (forward-1, forward-2, reverse-2, etc.). I know this question is old, but hopefully someone else who stumbles here will find this answer helpful.