In Article <47f0na$eqe at nntp3.u.washington.edu>,
joe at evolution.genetics.washington.edu (Joe Felsenstein) wrote:
>Are you inferring rooted or unrooted trees? From the picture, rooted.
>But parsimony, likelihood, and most distance matrix methods (such as
>Neighbor-Joining or Fitch-Margoliash) do _not_ estimate rooted trees.
>They estimate unrooted trees. And for three taxa there is only one possible
>unrooted tree, so all methods will give the same tree, which is always the
>right one, but the whole matter is therefore vacuous.
>>I can only interpret your posting as based on the total misconception
>that distance matrix methods infer rooted trees and do so by clustering
>algorithms, which put the closest pair together and then put the root
>Joe Felsenstein joe at genetics.washington.edu (IP No. 126.96.36.199)
> Dept. of Genetics, Univ. of Washington, Box 357360, Seattle, WA 98195-7360 USA
To Joe and other tree people,
How can you root a tree when a divergent gene is unknown?
For ex, in immunoglobulins, no one knows what would be a good
V-region "root" for v-region tree. Some say thy-1 , but its a shaky
v-region relative at best. I work on the evolution the immune system,
and am thinking of how to make a rooted tree with a new class of ig that
i recently cloned. I dont know if it is the "oldest" ig or if one or
several of its domains are older than the others. I make CLUSTALW
trees using NJ but would really like to do a rooted tree. Any suggestion?
Ralph M. Bernstein
Dept of Micro/Immuno
University of Arizona
Ph: 602 626 2585
Fx: 602 626 2100