Treeing partial sequences

Joe Felsenstein joe at GENETICS.WASHINGTON.EDU
Thu Mar 11 13:14:00 EST 1993


Doug Eernisse's posting reminds us that parsimony cannot be ruled out just
because we have partial sequences.  If we have some sites that are present  
in all sequences, both parsimony and distance methods (others too) will
be useable.  For distance methods it is clearly sufficient that there be
some sites in common between each pair of sequences -- they need not be
the same sites for different pairs.  For parsimony methods I suspect that
the minimum requirement will be that for each subset of four species, there
be some sites that are present in all of those four.

But no one has really worked out what these minimum requirements are.  It would
be nice if someone did.  Of course, the minimum requirements are far
short of what is needed to get an _accurate_ knowledge of the phylogeny.

The early postings in this thread dismissed parsimony too readily,
without asking whether the pattern of missing sites would be a problem for it.
I would wonder whether, in the original case, it would not be possible to
code the missing sites as "?", feed the sequences into either a
parsimony or a distance method (or others), and get a meaningful answer either
way.


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



More information about the Mol-evol mailing list