Modeling Human Evolution—To Tree or Not to Tree?
- Department of Pathology, Department of Biometry and Genetics, Stanley S. Scott Cancer Center, Neuroscience Center of Excellence, Louisiana State University Medical Center, New Orleans, Louisiana 70112 USA
This extract was created in the absence of an abstract.
Underhill et al. (this issue) report 19 new Y chromosome markers from a survey of 718 human genomes and use these data to construct a gene genealogy. According to the data, most of these biallelic markers were restricted to a few populations in specific regions of the world, but a few occurred at varying frequencies in all populations sampled, implying that they are older mutations. Underhill and colleagues constructed a genealogy of these polymorphisms using a novel combination of parsimony methodology and marker frequencies in 10 extant regional samples. In the discussion that follows, we distinguish the separate histories of chromosomal regions, individuals, and populations, as the interpretation of the branching process implied by a tree-like relationship between descendant nodes (the point at which branching occurs) becomes less clear as the level of analysis moves from chemical residues to aggregate populations. A clearer understanding of aggregate population structure and history may also be obtained using other conventional methods that do not impose a bifurcating process on the data. We illustrate one such method by reanalyzing the data reported by Underhill et al. (this issue) with principal components and plotting the principal coordinates of the populations (Cavalli-Sforza et al. 1994).
General Issues and Methods in Phylogenetic Inference
The branching pattern or topology of a network provides a graphical description of the mutation process that differentiates chromosomal haplotypes, provided that reversions of mutations to the ancestral states are rare. When the ancestral states of the mutations have been determined, networks may be rooted in a time dimension and are then considered trees. Trees can be constructed from either discrete character data or frequency data, and in either case the primary concerns are whether the topology and branch lengths are correct. Numerous tree-building methods have been proposed (e.g.,Sneath and Sokal 1973; Nei 1987 …
