Exploring linkage disequilibrium

Abstract

Linkage disequilibrium (LD, association of allelic states across loci) is poorly understood by many evolutionary biologists, but as technology for multilocus sampling improves, we ignore LD at our peril. If we sample variation at 10 loci in an organism with 20 chromosomes, we can reasonably treat them as 10 ‘independent witnesses’ of the evolutionary process. If instead, we sample variation at 1000 loci, many are bound to be close together on a chromosome. With only one or two crossovers per meiosis, associations between close neighbours decay so slowly that even LD created far in the past will not have dissipated, so we cannot treat the 1000 loci as independent witnesses (Barton). This means that as marker density on genomes increases classic analyses assuming independent loci become mired in the problem of overconfidence: if 1000 independent witnesses are assumed, and that number should be much lower, any conclusion will be overconfident. This is of special concern because our literature suffers from a strong publication bias towards confident answers, even when they turn out to be wrong (Knowles). In contrast, analyses that take into account associations across loci both control for overconfidence and can inform us about LD generating events far in the past, for example human/Neanderthal admixture (Fu et al.). With increased marker density, biologists must increase their awareness of LD and, in this issue of Molecular Ecology Resources, Kemppainen et al. () make software available that can only help in this process:LD na allows patterns of LD in a data set to be explored using tools borrowed from network analysis. This has great potential, but realizing that potential requires understanding LD.