What is a typical value of the selection coefficient s?

Question

I am trying to gain an understanding of the real world effects of natural selection from the equations, especially comparing it with drift. However I have been unable to find any examples which give an estimated value of the selection coefficient for a population considering a particular trait. Could anyone link me to some studies that give an actual value for a selection coefficient? Thank you in advance!

Answer 1

This is a very fundamental and important question! But first, be cautious not to confound the Fitness effect of segregating alleles with the mutational effects.

Fitness effect of segregating alleles vs mutational effects

One must not confond the distribution of fitness effect of segregating alleles in a population with the distribution of mutational effects. The distribution of mutational effects will always be narrower in comparison to the distribution of fitness effects of segregating alleles as very deleterious mutations won't stay long in the population and very beneficial mutation will sweep to high frequency very quickly so that it is unlikely to see them.

I am assuming you are interested in the distribution of mutational effects. Below are some papers and generalities that will be of interest to you. Some data come from the quantitative genetics literature and some from classical empirical population genetics. I also talk about mutation rate as these two subjects are highly related typically at determining mutation load.

Short answer

A typical deleterious mutation has a mutational effect of $s=0.1$ or $s=0.05$. In theoretical studies, authors often used a constant value of $s$ as being equal to one of these two numbers.

Longer answer

Distribution of mutational effect

Generally speaking distribution of mutational effects (on both the deleterious and the beneficial side) are thought to approximatively follow a gamma distribution with an extra bump of probability density around lethal mutations. shape might take values around 0.5 or 0.05 while the scale parameters might be of the order 0.1 to 0.001, with typically an average around 0.05. See Halligan and Keightley 2009. I have seen (see Nemo's documentation for example) that the dominance coefficent $h$ is modelled as $h=-log(2 Exp[h])/Exp[s]$.

Mutation rate

Concerning the mutation rate, a reasonable approximation is to consider a deleterious mutation rate per diploid genome of 1. This is known as Drake's law (see this post). Drake's law appears to be an underestimate for humans (Keightley 2012), about right for Drosophila, (Haag-Liautard et al. 2007), C. elegans (Denver et al. 2004) and possibly non-human endothermic vertebrates (Baer et al. 2007), and probably an overestimate for many other organisms (Baer et al. 2007, Halligan and Keightley 2009).

Quantitative traits

If you are uneasy with basic quantitative genetics and the below notation, you should have a look at this post.

For quantitative trait, there is evidence that $\frac{Vs}{Vp}≈5$ on average (Kingsolver et al. 2001, Johnson and Barton 2005). If heritability is $\frac{1}{3}$ (as it rarely significantly different from), then $Vs=7.5$. Empirical estimates of the mutational heritability $Vm$ suggests that $\frac{Vm}{Ve} = 0.001$ (Houle et al. 1996)