Can positive selection drown out neutral evolution? That's what John Hawks claims in response to my post on accelerated evolution. Hawks points out that, rather than looking at the neutral fixation rate (which is equal to the mutation rate, u), we should be more interested in the average time to fixation of a neutral mutation (the product of 4 times the effective population size, 4Ne). On the time scale he's looking at (40,000 years), neutral evolution shouldn't really matter because Hawks says that the Ne=100,000, which makes 4Ne>40,000.
Therefore, the majority of new mutations that arose in the past 40,000 years and have fixed (or nearly fixed) in a population should be either beneficial or linked to beneficial mutations. This allows for neutral evolution, however it's nearly all driven by hitchhiking upon the back of adaptive evolution. But this hypothesis hinges on the assumption that Ne>10,000; this is brought up at Popgen Ramblings in another post critical of the Hawks et al. paper. There's an important distinction between Ne (the effective population size) and the census population size. Ne is a parameter within population genetics theory that allows us to treat a population as ideal, but it is greatly influenced by the historical demography of a population (especially drastic decreases in population size). So much so that's it's not practical to estimate Ne from current population sizes or archaeological data -- Ne can only be estimated from population genetic data. The post at Popgen ramblings is skeptical of Hawks et al's estimate of Ne.
Finally, Hawks paints me as an anti-adaptationist in the tradition of Stephen Jay Gould. This is reminiscent of Larry Moran's bit where he claims to be a pluralist in the midst of the stark-raving-mad adaptationists. Only the roles are reversed. This time I'm the damn dirty neutralist, whereas I was previously labeled as the gullible adaptationist. I'm having a hard time keeping track of which team I'm on. Maybe that makes me the true pluralist.
- Log in to post comments
Nah, you're not a dirty anti-adaptationist! All these labels are nonsense; all that is important is understanding the math involved -- something Gould never really seemed very interested in. The problem with purely verbal arguments is that there is no scorekeeper: it's like Olympic ice dancing, or something.
As for effective population size, this is a long topic on which I have contributed some small things to the literature. Anyone who thinks that an effective size of 10,000 is plausible for humans within the last 40,000 years must explain how it is consistent with the pan-Old-World distribution (including Australia, Siberia, Southern Africa, etc.) of humans during that time. Remember that population subdivision and isolation tend to increase the effective size relative to census size. I have published on a number of mechanisms that would decrease effective size relative to census size; I don't think any of them are credible for the last 40,000 years, but please don't take my word for it, look up the literature.
Reasonable estimates put the census size in the millions by 40,000 years ago. Supposing that those hunter-gatherers had a population structure like recent ones, with approximately the same longevity, then Ne would have been on the order of half to one-third of census size. I can imagine other factors (excluding pseudohitchhiking, since that falls under the selection explanation) reducing effective size, but these must have become less potent toward the present. I propose 100,000 as an absolute minimum, and I expect that effective size was even larger -- and I am supported in that by genetics, which seem to show expansions to 100 times or more the previous Ne of 10,000.
A small follow-up to John's comment, which he probably already knows... Effective population size is only increased via subdivision if there is "local" population regulation (i.e., each group has equal offspring productivity); when there is variation in offspring production among groups in a subdivided population, then Ne is lower than the census size. The argument about a higher Ne hinges on complete isolation among demes (which makes the effective size infinite) but I'm not sure if this is a realistic scenario for ancestral humans.
That said, overlapping generations with extended reproductive lifespan will tend to increase Ne and bring it more in-line with N.
Finally, I just want to say that Hawk's treatment of the effective population size of ancestral humans is one of the most rigorous and considered treatments out there.
Rich Lawler said "The argument about a higher Ne hinges on complete isolation among demes". No, it really doesn't. Even a smidgeon of subdivision increases Ne: a decent approximation is that it multiplies Ne by (1+Fst) or divides it by (1-Fst), something like that.
The dazzling inconsistencies between census and effective size in the textbooks, as when one male fertilizes 10,000 eggs, don't really apply to humans much that we know of.
As I understand it, the relationship between population subdivision and Ne depends on whether the subdivided population experiences hard or soft selection. If groups within a subdivided population all produce equal numbers of offspring (soft selection), then Ne = N/(1-Fst). But if groups produce different numbers of offspring (hard selection) then Ne = N/(1+Fst). To me, hard selection seems more realistic for a subdivided population, given that groups likely differ in numbers of males/females and range over different habitats.
Thus, if you make the argument that population subdivision causes Ne to be greater than N, you need to make the additional assumption that offspring production among groups is equal--this is not likely to be the case for a majority of species.
But I do agree: the argument doesn't hinge on complete isolation as I previously wrote (in fact, I'm not really sure why I wrote that--it made sense to me at the time but it doesn't now), but given that among-group productivity is likely to be unequal for most species, it seems there will be few cases in nature where population subdivision leads to Ne > N.
I hate to sound like Alan Templeton but when we start talking about selection and population structure we have to specify what we mean by Ne. My habit is to think not of Ne but of the reciprocal, 1/Ne, which is the hazard of coalescence of a pair of genes.
In a subdivided population there are two cases--pairs of genes in the same deme and pairs in different demes. Initially (going backwards) the coalescence rate is higher in a subdivided population because coalescences occur within demes. Later, i.e. earlier going backwards, the rate slows way down because genes have to move between demes first. At this point Ne in my sense doesn't have any meaning at all.
What the formulae we are tossing around give, I think, is the time to the overall coalescence of a whole sample of genes. With hard selection going on it is of course faster in a subdivided population because one deme comes to be the whole population. In the familiar neutral case though subdivision slows the whole process down and makes for a deeper tree.
This right?? Henry
Is estimation of selection coefficients dependent on assumptions on how subdivided (or not) the population in question is?
If a population was highly subdivided but the model assumed that it was not, wouldn't that model result in overestimations of the strength of selection? I might be wrong.
It seems to me that the term "neutral mutation" is central in this discussion. I am wondering whether a masking effect by Heat shock proteins like Hsp90 (as mentioned in my blog http://sciphu.wordpress.com/ and described in: Hsp90 as a capacitor for morphological evolution Nature 396, 336 - 342 (26 Nov 1998)) is ever taken into account when considering evolution rate, - is it possible that some of these "neutral" mutations aren't neutral at all, but rather deleterious or beneficial mutations masked by a heat shock protein ?
I am in over my head when it comes to in-depth analysis of population genetics data, but still, to me, - the action of mutation masking Hsp's (if this is truly a valid evolutionary phenomenon) may seem to bridge the two views as well as solving a lot of other controversies surrounding the rate of molecular evolution vs. phenotypic/morphological evolution.