Biology Reference
In-Depth Information
4.3.2. Cladistic maximum parsimony (CMP) and character
compatibility (CC) methods
CMP methods search for the MP (maximum parsimony or most parsimo-
nious) tree(s), which minimizes the sum
S
of absolute dissimilarities between
all pairs of adjacent nodes in the tree.
o
Another way of explaining the CMP
method is that it creates trees in which the states of all characters are defined,
even those at internal nodes. The trees which are consistent with the small-
est sum
S
are considered the most parsimonious and are therefore retained.
To find the most parsimonious tree, CMP methods continually
compare hypotheses about the possible states of ancestral nodes. They
proceed by progressively looking for synapomorphies, thus indirectly
minimizing the total degree of homoplasy. Since the methods need to
find synapomorphies, it becomes evident that a character with states
differing from one group of taxons to another group is more informative
than a character which has the same state throughout all taxons, or all
taxons minus one. The differences between taxons are methodologically
more important than the similarities.
p
CC methods proceed in roughly the same manner, but they look for
the global MP tree(s) that can account for the largest clique (or set) of
characters without having to resort to any hypothesis of homoplasy.
Common parsimony programs include PAUP*,
2
TNT,
3
and MEGA.
4
Parsimony methods are generally much slower than NTP methods,
mainly due to the nature of the heuristic methods used.
4.3.3. CMP common artefacts
Since dissimilarities are not additive, a cladophylogram generated by a CMP
or CC method is not necessarily a good indicator of relative evolutionary
o
Note that a node is only adjacent to its direct ancestor and to its direct descendants.
Since the dissimilarity in NTP methods is calculated between pairs of terminal taxa,
which can never be adjacent, the values of an NTP dissimilarity semimatrix are never
used in CMP.
p
Computer algorithms proceed differently to reconstruct CMP trees. They simply
calculate the length of all possible trees and retain the best results. If heuristics are
used, only the subset of trees most likely to contain MP trees is calculated.
