Maximum likelihood distances for aa data

Joe Felsenstein joe at evolution.genetics.washington.edu
Mon Aug 28 02:00:40 EST 1995


In article <carmean-2408951714130001 at creslab2.biol.sfu.ca>,
Mary Berbee <carmean at sfu.ca> wrote:
>We would like to to generate maximum likelihood distances (branch
>lengths), with confidence intervals, for protein sequences using a given
>phylogeny (user tree).

If I understand the objective, the User Tree mode of PHYLIP's DNAML program
would do just that.  The tree would have to be in unrooted form, but the
program RETREE could make it so if it were not. 

Be warned that the confidence limits are underestimates, however, as they
keep all other branch lengths constant while changing one of them.

A tedious but workable way to get better confideence limits would be to 
put in a tree as a user tree with no branch lengths on it except on one
branch.  Try a bunch of different values for that branch length, and
select the option that asks the program to use the branch lengths that are
read in (in that case there is just one of them).  All the branch lengths
that give likelihoods within 1.92 of the maximum are in the confidence limit.

You have to do that branch by branch.

-----
Joe Felsenstein         joe at genetics.washington.edu     (IP No. 128.95.12.41)
 Dept. of Genetics, Univ. of Washington, Box 357360, Seattle, WA 98195-7360




More information about the Bio-soft mailing list