protein parsimony and significance

Joe Felsenstein joe at evolution.genetics.washington.edu
Tue Mar 14 07:35:57 EST 1995


In article <1995Mar13.183053.35578 at ac.dal.ca> aroger at ac.dal.ca writes:
>In essence I wish to do 
>protein parsimony without any stepmatrix and decide between the
>best tree and some alternative trees.

I am not sure what that means.  Do you want to do parsimony allowing
1 step for a change from any amino acid to any other?  Why is that
at all reasonable?

> I'm relatively poor at
>math so I haven't been able to understand the methods described by
>Kishino and Hasegawa, (1989), 29:170-179 J. Mol. Evol. and figure
>out how they apply to protein sequence data.
>
>Any simple answers will be appreciated!

The Kishino-Hasegawa-Templeton test is probably your best bet here.  Kishino
and Hasegawa did the likelihood version with statistical justification.
Alan Templeton invented the original test with parsimony.  His 1983 paper
may not be clear but you might find some additional material in my
1985 paper in Systematic Zoology on confidence limits for phylogenies with
a molecular clock.  The description of the test there as a paired sites
test is simpler (and will be helpful even if you can't assume a clock).

Basically it is just a nonparametric test for two samples with paired
observations, testing whether the means (and thus the sums too) of the
two samples are different.  The sites are regarded as a sample from a
distribution.  Statistics books covering nonparametric statistics can be
consulted for details.

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



More information about the Mol-evol mailing list