Estimating divergence times when branch lengths vary

Thorsten Burmester thorsten at erfurt.thur.de
Tue Jun 15 10:25:27 EST 1999


Dear Evolutionists,

Let us consider a phylogenetic tree of a superfamily of orthologous
proteins. There are three or more distinct protein families, each with a
rather constant evolution rate. I.e., the molecular clock assumption
works pretty well within each family. However, between them there is a
rate variation of a factor >2.

                       +------------------A
                a      |
            +----------+2   +---------------B 
            |          +----+
            |               +--------------C 
     +------+ 1
     |      |      +--------L
     |      |  b   |
     |      +------+3 +-----M 
     |             +--+
 +---+                +-----N 
     |  
     |       +-------X
     |       |
     +-------+  +-----Y 
             +--+
                +-----Z 



After estimating the divergence times within the families (here A,B,C;
L,M,N; X,Y,Z), I am now interested whether you could do the same for the
internal nodes. Let us consider node #1. When "knowing" the time of the
branching event of node #2 and branch length a, may I extrapolate to
node #1, or are there problems I need to consider? I have used node #3
and length b as a control to see whether I calculate the same time for
node #1.

Some more questions: 

1.) Do the different phylogenetic programs assign the branch lengths
"correctly" to these internal branches (like a)? Are there different
assumptions (apart from the general phylogenetic method used) that are
implemented in the programs to calculate these lengths? I have the
slight feeling that some programs (especially PAUP parsimony) tend to
underestimate them.

2.) Is there any statistical test to verify the estimated divergence
times, i.e. to say "a divergence time of X MYA can be rejected with P <
0.001" or s.th. like this.

3.) Hints to papers and computer programs are very welcome.

Thorsten

-- 
Thorsten Burmester
thorsten at erfurt.thur.de




More information about the Mol-evol mailing list