Calcium-Binding Protein Protocols Calcium-Binding Protein Protocols

Multiple Sequence Alignment 239
preparing the alignment is certainly worth the effort. During the course of such
optimizations, it is also worthwhile considering the incorporation of external
information garnered from biochemical or structural studies of the proteins in the
alignment.
6. Three basic concepts exist to reconstruct the divergent evolution of a set of
sequences: (a) parsimony (53,54); (b) distance (55); and (c) maximum likelihood
methods (56). Parsimony methods try to reconstruct phylogenies by exploring
the concept of minimum mutation. Distance methods are aimed at exploring a
matrix containing all pairwise distances of a set of multiply aligned sequences.
These methods also try to reconstruct the past using a minimalist approach; i.e.,
using as few evolutionary changes as possible. Maximum-likelihood methods
attempt to construct the most probable tree based on the sequence data and a
specific stochastic model of evolution. The additional information that can be
expressed in the model, such as weighting of functional similarity and amino
acid importance and the nature of insertions/deletions (e.g., ref. 10), can optimize
the connectivity and branch lengths of the resulting tree. The package
MOLPHY (57) is a speed-optimized maximum likelihood method that could be
considered if evolutionary information is the most important analysis. In this
chapter, we will restrict ourselves to distance-based methods as they are quick
and can be applied easily onto the sequence data.
The evolutionary relationships of a subject set of sequences are normally
depicted in a tree. A tree is a special case of a connected graph where travel from
each node to any other is possible through edges (branches) by only one path
between any two such nodes. A tree contains interior and exterior (terminal)
nodes. Normally, the input sequences are contemporary and referred to as the
operational taxonomic units (OTUs). They correspond with the exterior nodes of
the evolutionary tree, whereas the internal nodes represent ancestral sequences
that must be guessed from the OTUs and the tree topology. The length of each
branch connecting a pair of nodes may correspond to the estimated number of
substitutions between two associated sequences. The minimal evolution hypothesis
is that the “true” phylogenetic tree is the rooted tree (i.e., contains a node
ancestral to all other nodes), which has the shortest overall length and thus comprises
the lowest cumulative number of mutations.
Distance methods derive a tree from a distance matrix, in which approximations
are stored of all pairwise evolutionary distances between the tree constituents
(i.e., the sequences). Distances can be obtained from sequence identities
(55) or pairwise sequence alignment scores (58). Normally, an agglomerative
cluster criterion is then used to construct the phylogenetic tree, reflecting the evolutionary
information in the best possible way. Many clustering criteria have been
introduced over the years and each has an underlying assumption of evolutionary
dynamics. The first cluster method used in molecular sequence phylogeny was
the UPGMA or group-averaging method (38). It takes the average value over all
intergroup distances to measure the evolutionary distance between two groups of
sequences and has the underlying assumption of identical mutation rates in all