242
isolated and digested with restriction endonucleases, which cleave the DNA at specific
sequences (usually 6 nucleotides in length). This produces DNA fragments of varying lengths,
which can be size fractionated by agarose gel electrophoresis. DNA fragments containing
sequences homologous to specific genes can be visualized by hybridization with radiolabelled
gene probes. The number and size of these DNA fragments should be nearly identical in very
closely related organisms, while divergent species will display different DNA restriction
fragment patterns (restriction fragment length polymorphisms). Using a variety of restriction
endonucleases and hybridization probes, genetic relatedness can theoretically be estimated
from these polymorphisms. However, several sources of error inherent in RFLP data render
this technique less useful than direct sequence comparisons for phylogeny reconstruction
(Swofford and Olsen, 1990).
In contrast to the above approaches, direct acquisition of macromolecular sequence data allows
quantitative comparisons of sequences from homologous genes or proteins. Once obtained,
archived sequences provide a reference data base which can be directly and immediately
compared to newly acquired sequences. The data are first analyzed by aligning homologous
nucleotides or amino acid positions of sequences to be compared. Highly conserved regions
of sequence can serve as convenient landmarks in the initial alignment process. After the
homologous sequence positions are juxtaposed, they are compared and analyzed by any of a
variety of available methods (see below). The analyses can be computationally intense because
of the large amount of data compared, and the statistical algorithms used to infer phylogenetic
tree topology and estimate error. Hence, molecular phylogenetic analyses usually require
extensive use of computers. A number of numerical methods for inferring phylogenies from
molecular data are currently in use. The methods are rapidly improving, as computer
algorithms become more sophisticated and computer facilities more readily available. The
major approaches now used for estimating molecular phylogenies are cluster and distance
matrix methods, parsimony analysis, and maximum likelihood methods (Swofford and Olsen,
1990). Distance matrix methods calculate a similarity value (S) between each pair of
homologous sequences in a given data set. A pairwise evolutionary distance value can then
be derived from sequence dissimilarity (l-S), which takes into account the possibility of
multiple mutations at any given nucleotide position. These pairwise distance values are
assumed to be strictly additive in the phylogenetic tree construction. The best tree is that
isolated and digested with restriction endonucleases, which cleave the DNA at specific
sequences (usually 6 nucleotides in length). This produces DNA fragments of varying lengths,
which can be size fractionated by agarose gel electrophoresis. DNA fragments containing
sequences homologous to specific genes can be visualized by hybridization with radiolabelled
gene probes. The number and size of these DNA fragments should be nearly identical in very
closely related organisms, while divergent species will display different DNA restriction
fragment patterns (restriction fragment length polymorphisms). Using a variety of restriction
endonucleases and hybridization probes, genetic relatedness can theoretically be estimated
from these polymorphisms. However, several sources of error inherent in RFLP data render
this technique less useful than direct sequence comparisons for phylogeny reconstruction
(Swofford and Olsen, 1990).
In contrast to the above approaches, direct acquisition of macromolecular sequence data allows
quantitative comparisons of sequences from homologous genes or proteins. Once obtained,
archived sequences provide a reference data base which can be directly and immediately
compared to newly acquired sequences. The data are first analyzed by aligning homologous
nucleotides or amino acid positions of sequences to be compared. Highly conserved regions
of sequence can serve as convenient landmarks in the initial alignment process. After the
homologous sequence positions are juxtaposed, they are compared and analyzed by any of a
variety of available methods (see below). The analyses can be computationally intense because
of the large amount of data compared, and the statistical algorithms used to infer phylogenetic
tree topology and estimate error. Hence, molecular phylogenetic analyses usually require
extensive use of computers. A number of numerical methods for inferring phylogenies from
molecular data are currently in use. The methods are rapidly improving, as computer
algorithms become more sophisticated and computer facilities more readily available. The
major approaches now used for estimating molecular phylogenies are cluster and distance
matrix methods, parsimony analysis, and maximum likelihood methods (Swofford and Olsen,
1990). Distance matrix methods calculate a similarity value (S) between each pair of
homologous sequences in a given data set. A pairwise evolutionary distance value can then
be derived from sequence dissimilarity (l-S), which takes into account the possibility of
multiple mutations at any given nucleotide position. These pairwise distance values are
assumed to be strictly additive in the phylogenetic tree construction. The best tree is that