This post was written by Fabian Staubach.
The neutral theory of molecular evolution assumes that adaptation is rare and that the effect of adaptation on linked variation, the so-called hitchhiking effect, typically has only little influence on the dynamics of molecular genetic variation. Because of this assumption, it is widely assumed that in most natural populations, hitchhiking can be neglected, or at least reasonably well approximated by the introduction of effective parameters, such as an effective population size. But if molecular adaptation is in fact common, then the assumption may be violated, and we should worry whether population genetic methods and estimates of evolutionary parameters obtained from them are robust to frequent hitchhiking.
In their paper “Frequent adaptation and the McDonald-Kreitman test” (PNAS, 2013), Philipp Messer and Dmitri Petrov investigate this question for one of the key population genetic methods — the McDonald-Kreitman (MK) test. This test forms the basis of most commonly used approaches to measure the rate of adaptation from population genomic data and has been used to argue that in some organisms, such as Drosophila, the rate of adaptation is surprisingly high.
The MK test can substantially underestimate the true rate of adaptation
Messer and Petrov employ their powerful forward simulation software, SLiM (see here), to simulate the evolution of entire chromosomes under a range of parameter values relevant to humans and other organisms, and apply various forms of the MK test to the population genomic data resulting from their simulations. They then study how accurately these methods re-infer the true evolutionary parameters in the simulations. Strikingly, they find that the MK test can substantially underestimate the true rate of adaptation. For instance, they present scenarios where 40% of the amino acid changing substitutions were in fact strongly adaptive in the simulations, while other population parameters resembled those commonly inferred for human evolution, yet the standard MK estimates yield that none of these substitutions were actually adaptive. Fortunately, Messer and Petrov propose a way to avoid these problems by using a simple, asymptotic extension of the MK test.
Figure: Illustration of the asymptotic MK estimation of the rate of adaptive substitutions : The standard MK approach assumes that all polymorphisms (non-synonymous and synonymous) are neutral. This assumption is likely violated for low frequency polymorphisms, as some of these are likely to be deleterious. The assumption should hold for very high frequency polymorphisms, because they are very unlikely to be deleterious. The asymptotic MK approach uses this fact by looking at the estimated rate from different frequency classes of alleles, and extrapolating to x=1, where the rate is expected to have asymptoted.
The bigger claim of this straightforward and easy-to-read paper is that the effects of linked selection cannot be simply swept under the rug by introducing effective parameters, such as effective population size or effective strength of selection, and then using these effective parameters in formulae derived from the diffusion approximation under the assumption of free recombination.
Quantifying known biases
Surely, this paper will ruffle some feathers. Some people will argue that these problems have been know for a while in theory. Yet despite this, the vast majority of studies that continue to appear in the literature still pay only cursory lip service, if anything, to these issues. Presumably, this is because it is not well understood analytically to what extent linkage effects affect population genetic estimates, and Messer and Petrov therefore do an important job in quantifying these biases. Hopefully this will help focus the community’s attention to spend some time figuring out how to modify commonly used approaches to place them on a more solid foundation.
Citation: Messer, P. W., & Petrov, D. A. (2013). Frequent adaptation and the McDonald-Kreitman test. Proceedings of the National Academy of Sciences of the United States of America, 110(21), 8615–20. doi:10.1073/pnas.1220835110