The study of structured populations — new hope for a difficult and divided science

Abstract

Population genetics can be used to study the history of natural populations. However, it is a difficult science because natural populations have complex geographies and histories. With the advent of DNA-sequence-based data sets drawn from natural populations two main schools of study developed: the phylogeographic approach, which uses the data to estimate the evolutionary tree, or gene tree, then attempts to interpret the history of the populations from which the samples came; and the summary statistics approach, which is an outgrowth of mathematical population genetics and proceeds by mathematically fitting specific population-genetic models to the data. The phylogeographic approach has the advantage of not being constrained by specific models, and lends itself to exploratory types of analysis. However, it is highly dependent on gene-tree estimates, which are often incorrect. This method can be misleading if investigators focus on just a single gene or stretches of tightly-linked sequence, such as mitochondrial DNA, and overlook the large stochastic variance that arises among genes in populations. Summary-statistic approaches can be mathematically sophisticated and provide ways to compare models and assess the sources of variance in the process that gave rise to the data. However, these methods are often highly constrained by the available models and are difficult to apply if investigators have little knowledge of the locations and boundaries of populations in nature. Also, they do not usually take full advantage of all of the information that is available in the data. In recent years, a new family of methods has begun to offer the advantages of the phylogeographic approach, using all of the information in the data and allowing diverse models to be considered, together with the mathematical sophistication of the summary-statistics methods. These are probabilistic methods in which gene trees have a role, but in a framework in which they are used strictly in conjunction with their probability. As these methods continue to develop, they offer the promise of increased flexibility and applicability to a wide range of questions in the history of populations.