Dear Sergei, Art, & Simon,

I constructed a tree in PHYML and fit a MG94xHKY85 model to it in HyPhy. HyPhy inferred several internal nodes that are duplicates of each other, i.e., the sequences are identical. Not all of them are obvious polytomies. In other places the internal nodes have the same sequences as tips (sampled taxa).

When calculating dN/dS for each branch with this model, I found that the branches terminating in identical sequences have strange patterns of dN/dS differences. I would expect that branches would either share the same dN/dS, or that one would reflect divergence from the ancestor and the other would be 0/0 = NaN. Instead, all kinds of combinations appear, including some w/o NaN:

inf, NaN, 1, NaN
[high real #], inf
inf, NaN
0, 0
1.5, NaN
0.4, NaN
(etc.)

What causes these differences?

Thanks,
Sarah

Hi, Sergei,

I'll email everything to you. Thanks.

Sarah