Genetical Research

Individual organisms as units of analysis: Bayesian-clustering alternatives in population genetics

a1 Department of Genetics, University of Georgia, Athens, GA 30602, USA

Article author query
mank je   [PubMed][Google Scholar] 
avise jc   [PubMed][Google Scholar] 


Population genetic analyses traditionally focus on the frequencies of alleles or genotypes in ‘populations’ that are delimited a priori. However, there are potential drawbacks of amalgamating genetic data into such composite attributes of assemblages of specimens: genetic information on individual specimens is lost or submerged as an inherent part of the analysis. A potential also exists for circular reasoning when a population's initial identification and subsequent genetic characterization are coupled. In principle, these problems are circumvented by some newer methods of population identification and individual assignment based on statistical clustering of specimen genotypes. Here we evaluate a recent method in this genre – Bayesian clustering – using four genotypic data sets involving different types of molecular markers in non-model organisms from nature. As expected, measures of population genetic structure (FST and ΦST) tended to be significantly greater in Bayesian a posteriori data treatments than in analyses where populations were delimited a priori. In the four biological contexts examined, which involved both geographic population structures and hybrid zones, Bayesian clustering was able to recover differentiated populations, and Bayesian assignments were able to identify likely population sources of specific individuals.

(Received January 20 2004)
(Revised August 20 2004)

c1 e-mail: