Outgroups and polarized character states

Joe Felsenstein joe at evolution.genetics.washington.edu
Sat Nov 8 11:20:28 EST 1997

In article <63o6sg$480 at net.bio.net>,
Guy A. Hoelzer <hoelzer at med.unr.edu> wrote:
>Andrew has correctly described the way that outgroup analysis was
>originally intended to work.  However, it is rarely applied in this way by
>the commonly used computer programs, such as PAUP.

One should just add that this is not an oversight.  For example, when
a site has G at all ingroup states, and A in the outgroup, the
ancestral state reconstructed by parsimony could be either A or G.
Using the outgroup state as the ancestral state is always one of the
most parsimonious state reconstructions, but there can be others.
The authors of most programs were well aware of this and chose not to
simply usee the outgroup state, but to show the fuller information.

The phrase "polarize" used in this thread is perhaps unfortunate.  If there
are (say) two states, 0 and 1, we could imagine two cases:

1) We want to allow symmetric change between 0 and 1, so that the
model is, in effect  0 <--> 1.  In this case (Wagner Parsimony) there is no
polarization, as having reached 1 one can go back to 0.

2) We think that change has been unidirectional, but do not know which
state was ancestral.  The changes are either all 0 --> 1 or all 1 --> 0.
In my parsimony programs I have called this Unordered Camin-Sokal Parsimony.
It is simply Camin-Sokal Parsimony with the ancestral state unknown.
This may be the case that the original posting was thinking of when
using the term "polarize", for having chosen the ancestral state by
(in this case) considerations of parsimony, we have chosen among the
two directions.  If we choose 1 as the ancestral state, we then specify
that all changes must be interpeted as 1 --> 0 with no reversals

Joe Felsenstein         joe at genetics.washington.edu
 Dept. of Genetics, Univ. of Washington, Box 357360, Seattle, WA 98195-7360 USA

More information about the Mol-evol mailing list