Distance methods, along with character methods and likelihood methods,
form one of the three major approaches to estimating phylogenies from
DNA or protein sequence data. Distance methods all start by
estimating the evolutionary distances between pairs of extant species,
and then seek to build phylogenies that fit the estimated distances as
closely as possible. We survey the leading methods, including the
popular neighbor-joining algorithm, various least-squares methods, and
minimum evolution methods. We consider the algebraic basis common to
these methods, and show how the slight differences in approaches lead
to significant differences in computational time requirements.
Finally, we consider how the various methods perform when tested
against simulated data generated under biologically reasonable
conditions.
|