Proof or counter-proof... this is how to do it.

Mary K. Kuhner mkkuhner at phylo.genetics.washington.edu
Sat Jul 15 08:03:15 EST 1995


I think there are two important points being missed here:

(1)  It is not necessary to assume a molecular clock in order to make
phylogenies.  Both the method of maximum likelihood and various distance
methods (using distances that have been appropriately corrected for
multiple hits) can very successfully recover the phylogeny even if some
lineages evolve ten times as quickly as others (as shown by several
groups' computer simulations).

(2)  When you make two phylogenies using different data sets, and they
do not agree, there are a number of possible explanations.  In my
opinion the most likely one is that neither phylogeny is precise.
Phylogenies estimated from data are estimates with error.  Two
error-prone estimates may differ even if there is nothing particularly 
wrong with either one.  This can be tested with various sorts of
bootstrapping.  It's premature, in my opinion, to draw any far-ranging
conclusions from mismatch between phylogenies until you first establish
that they are not within each others' range of error!

It is very usual for two phylogenies made from computer simulated data
generated under *exactly* the same assumptions to disagree, especially
if the tree includes a large number of sequences--unless the sequences
are extremely long and thus very informative.  The models used to
analyze the data may be 100% perfect with simulated data, but the answer
still shows stochastic error.

Mary Kuhner mkkuhner at genetics.washington.edu



More information about the Mol-evol mailing list