Tree construction: parsimony versus distance-based

VICTOR B. STRELETS STRELETS at SCRI.FSU.EDU
Tue May 24 09:58:00 EST 1994


Hi,

Both parsimony and distance-based methods are completely wrong
and can be used only for approximate estimations of tree
topology. They are based on homology (similarity) evaluations.
In the best cases, for SUBsequences.

But ALL THESE LAST YEARS demonstrated importance of absolutely
another information. REAL tree should take into account 
tertiary fold for subsequences under comparison rather
their word-style similarity. Plus information about evolution
history (to compare only subsequences which belongs to one and
the same evolutionary unit, which remained undisturbed or undergo 
commonly-reasoned changes in the course of recent events).
Standard (for now) and well known inverse folding problem which 
is often ignored by forestry division.

And optimal way to construct trees definitely consist of
 - preliminary construction of tree topology approximation
   by means of mentioned in discussion old good methods
 - analysis of corresponding homologies on the base of
   structure- or evolution-based techniques
 - iterative (until dead) correction of final tree.
And you may only hope to have a good (REAL) result. Because
nobody knows how to reveal true evolutionary blocks.

Any discussions about parsimony or distance-based methods are
applicable only for multiple alignments (good fast food for 
experimentators). Implications to the evolutionary consequences
as well as attempts to find/explain correlations between
"phylogenetic" trees and archeological (real evolution) facts
seemed to be incorrect and in some cases even pseudo-scientific.

And even for multiple alignments, similarity comparisons does
not work well: remember need to exclude gap penalties, recent
findings of structure similarity without same at sequence level
(so many such regions are present in 3D-align database files
which contain these alignments) etc.

Nothing to discuss - take any method if you do alignments
or generally reconstruct ALL results way if you 
try to build something evolutionary.

STR





More information about the Bio-soft mailing list