In article <610o18$mgl at net.bio.net>,
John Clement <clement at mail.utexas.edu> wrote:
>I was wondering if there are methods available for determining if there
>are branches on an evolutionary tree where amino acid sequences have
>undergone signficant amounts of non-conservative change in their
>side-chain chemistry.
... [and using the Categories option of my PROTDIST program ...]
>In other words, if you have three branches of a tree, a, b, and c,
>estimate the distances using both values, and fit them to the tree. If
>l(a,0.1)/l(a,1.0) is much greater than l(b,0.1)/l(b,1.0) and
>l(c,0.1)/l(c,1.0), then that is evidence for a greater amount of
>non-conservative change on a than on b and c. Or perhaps one could
>weight the branch lengths by the proportion of change each represent in
>the tree. I. e., instead of l(a,0.1), use
>>l(a,0.1)/(l(a,0.1)+l(b,0.1)+l(c,0.1))
>>and so forth. Is this a legitimate approach?
Well, you can get numbers out that way, sure. But how do you test
whether they are significantly different from each other?
At the moment we do not have (in any program that I know of) a model in
which the amount of conservative versus nonconservative amino acid
substitution is allowed to vary by branch and be tested as to
whether it differs among branches. There might be some way you could
sort-of do it by boostrapping positions, and then comparing how often
l(a,0.1)/l(a,1.0) is much greater than l(b,0.1)/l(b,1.0) and l(c,0.1)/l(c,1.0).
--
Joe Felsenstein joe at genetics.washington.edu
Dept. of Genetics, Univ. of Washington, Box 357360, Seattle, WA 98195-7360 USA