Recombination of haplotypes leads to biased estimates of admixture proportions in human populations

Abstract

A population formed by genetic admixture of two or more source populations may exhibit considerable linkage disequilibrium between genetic loci. In the presence of recombination, this linkage disequilibrium declines with time, a fact that is often ignored when considering haplotypes of closely linked systems [e.g., Gm serum group (gamma globulins), HLA, and, more recently, restriction fragment length polymorphisms]. Recombination alters haplotype frequencies over time, and the haplotype-derived measures of admixture proportions from haplotype frequencies in generations following the admixture event become progressively more biased. The direction and extent of this bias can be predicted only when the history of admixture is known. Numerical illustration suggests that this bias is problematic whenever rt greater than 0.05, where r is the recombination rate between linked loci and t is the time (in generations) that has elapsed since the admixture event. In general, even the haplotype frequencies defined by multiple restriction fragment length polymorphism should be used with caution for admixture analysis. When recombination rates or the time since admixture are not precisely known, it is advantageous to consider each restriction fragment length polymorphism site separately for admixture analysis.