So far, we examined Craig’s argument that the mouflon sheep study – a study that traces a population of sheep descended from a single breeding pair – lends scientific credence to his assertion that humans also descend uniquely from a pair (rather than from a population of about 10,000 individuals, which is the current scientific consensus). As we have seen, Craig failed to recognize that the effects of selection can maintain genetic diversity (in the form of heterozygosity) in a population – and that this is not a failure of population genetics models, but rather an expected consequence of selection.
So, why is it that selection has worked to maintain heterozygosity across large regions of the genome in this population of sheep? And how can we be sure that similar effects are not similarly confounding our estimates of ancestral human population sizes? To address these questions, we’ll need to learn a bit about how heredity works in small populations over short timescales.
In a population founded by a single breeding pair, all progeny in the first generation will, of course, receive half of their chromosomes from this one male and half from the one female. Genetic differences between the original pair – i.e. different alleles that they carry on their chromosomes – will thus be inherited on these chromosomes. If we were to examine one genome region in one of the first generation progeny, we might see something like the following:
In this region, we see four locations where there are differences between the chromosomes inherited from the male and female. These allele differences can be represented with letters, as we have discussed. In this example, this sheep would be heterozygous at four locations in this region.
Now imagine that this sheep reaches breeding age and passes on its chromosomes. It would pass on these same groupings of alleles, unless there was mixing and matching of alleles through recombination (which you might recall as “crossing over” from high school biology). Crossing over may occur during the cell divisions that produce eggs and sperm, giving rise to new combinations of alleles:
While crossover events between any two genome locations are possible, in practice it can take many, many generations before two locations close together are recombined. What this means is that combinations of alleles will stay together in groups for long periods of time. In a population founded by only two individuals, the original allele groups from the founding parents will be broken up slowly over tens of generations through crossover events.
So, how does this affect influence selection and heterozygosity? It means that any location that is under selection will have a large genome region around it that will tag along for the ride, as it were. Let’s look at an example to illustrate this effect. Let’s return to the original combinations present in the first generation, but now consider what will occur if one location in this region is under selection to be heterozygous:
Suppose the “cee” location is under selection for heterozygosity. What this means is that animals that have both the “C” and “c” alleles will leave more progeny in the next generation than animals that are “CC” or “cc” (i.e. homozygous). In the absence of crossing over, however – and remember, crossing over in any given location is an infrequent event – selection for a “Cc” animal will at the same time select for animals that are heterozygous at the other three nearby locations. In this way, these other locations are also selected for heterozygosity, even though the real selection is acting only at the “cee” location. The other locations are merely “hitchhiking” because they are physically connected to the alleles under selection.
In a small population over a short timeframe (tens of generations) this effect can be quite pronounced. All it takes is a few locations in the genome to be under selection for heterozygosity, and large swaths of the genome will hitchhike along. This means that a large part of the genome will not act as though it is selectively neutral, even if the alleles themselves are neutral. For example, it may well be that the alleles at the “aye” location are neutral, meaning that the AA, Aa and aa combinations are all equally likely to leave progeny in the next generation. Based on this alone, there would be no disadvantage to this population to lose either the “A” allele or the “a” allele, with the resulting loss of heterozygosity. Moreover, should either allele be lost, it is lost for good – or at least until a mutation event re-creates it, something far too rare to be expected over a small number of generations. In our example, however, selection at the “cee” location also selects for heterozygosity at “aye” – and actively maintains both alleles at this location in the population over time. So, this is what is going on in the mouflon population – and this is why the authors find heterozygosity that exceeds the level expected from the loss of alleles not under selection. None of this is particularly surprising, of course, and is plainly laid out in the original research article. In the presence of selection, we expect this effect to occur.
Despite the confounding effects of selection, the effects on estimating the mouflon population size are not overly significant. The known ancestral population size is two; even taking the data at face value (assuming no selection) only raises the estimate to a maximum of eight individuals. For Craig to argue that this result somehow casts doubt on human studies is a flawed argument, for several reasons. First and foremost, ancestral human population size estimates are derived from many independent measures, not merely measures of heterozygosity, which is a rather simplistic measure. Secondly, we have full genome sequence data for tens of thousands of humans, and from this data we can directly observe that the short-term “hitchhiking” effects confounding the Mouflon study are not similarly confounding human studies. Thirdly, even if the confounding effects in the Mouflon study were somehow directly proportional to human studies (which for the foregoing reasons they are most certainly not), the end result would be an ancestral human population size of about 2500, not 2 as Craig requires.
So, the sheep study – though understandably tempting for apologetics purposes – in fact cannot be used to build a scientific case for humans descending uniquely from an ancestral couple. Those who attempt to do so, like Craig (and Reasons to Believe, on whom Craig is depending) merely demonstrate that they do not understand population genetics.
Next, we’ll look at an additional problem for Craig’s model: the data that indicate that some present-day humans descend in part from non-human species such as Neanderthals.