Stephen W. Schaeffer
sws4 at PSUVM.PSU.EDU
Thu Apr 28 07:42:02 EST 1994
On Thu, 28 Apr 1994 06:29:05 GMT, Mark Siddall wrote:
>In article <1994Apr26.185539.900 at honte.uleth.ca> std_dickout at hg.uleth.ca writes:
>>I am interested in information and/or opinions concerning the use of Maximum
>>Parsimony analyses and Distance analyses with respect to the treatment of
>>molecular data in the field of cladistics. Explanations dealing with
>>how these two methods differ from one another would also be appreciated.
>>Please mail me directly.
>>Thank you very much.
>Okay, hang on this is gonna take a bit of work.
>The reason it's called MAXIMUM parsimony is because reversals are allowed
>with the same rate as unitary changes (apomorphies) as are convergent
>changes (> unitary (i.e., occurring more than once)). That is to say all changes regardles of type are of the same weight.
>Parsimony applies a DIRECTION to the the difference between the states observed
>between taxa. That is, with 3 taxa one with state "a" the other two with
>state "b". Parsimony requires that one can specify whether "a" or "b"
>came first. If "a" came first then there's a single step uniting the two
>taxa with "b" as having descended from a common ancestor that had "b".
>If "b" came first, then there's nothing uniting the two with "b" and the one
>with "a" stands alone. So, only DERIVED changes are considered important to
>So... one has a choice of considering only a particular kind of differnece
>(i.e., directional; i.e., derived) as informative, or any kind of differnece
>(uninformtive of ancestry or informative of ancestry) as informative. Distance analyses are of the type that calculate absolute difference (irrespective
>of KIND of difference). Thus, by analogy, one might group crocodilians
>with saurians (lizards etc) on the basis of OVERALL SIMILARITY (ABSOLUTE
>DIFFERENCE or DISTANCE) in spite of the fact that most of the similarities with
>sauriance are ancestral. Whereas, the DERIVED characteristics suggest that
>they share a most recent common origin with birds.
>In short distance measures of any sort are phenetic.
>Sneath and Sokal in their original "Numerical Taxonomy: stated quite clearly
>in Chapter 9 that phenetic techniques are NOT phylogenetic and should not
>be contrued as such.
>Thus, if you want to classify organisms, phenetic or phylogenetic methods could be
>equally justified. BUT if you're interested in the pattern of evolutionary
>descent of the organisms, the most stalwart adherents have already told you tha
>that measures of similarity won't do it!
>Thus from an evolutionary point of view, phenetics, distance etc add nothing.
>Moreover, from a genetic evolutionary point of view, the most often used method
>Fitch-Margoliash Distance, has been criticised by many as being invalid
>including one of the authors of the method!!
Distance methods ARE based on an evolutionary model, the neutral model
of molecular evolution. This model assumes that nucleotide substitutions
accumulate in a clock-like manner over time because of mutation and genetic
drift. Thus, the fraction of nucleotide differences between two
sequences are an indirect measure of the evolutionary time back to a common
ancestor. One difficulty with comparisons of nucleotide sequences among
distantly related sequences is that a homologous nucleotide position may
change multiple times through evolutionary time. Distance measures are
corrected for multiple substitutions that may occur at nucleotide sites.
Cladistic methods do not take the multiple substitution process into
account. This is not a severe problem if the sequences that are compared
in a cladistic analysis are closely related.
Stephen W. Schaeffer
Institute of Molecular Evolutionary Genetics
The Pennsylvania State University
More information about the Bioforum