So I read Sanford et alâs 2018 paper where they attempt to reconcile the allele frequency spectrum (AFS) with a YEC model, and I have a few questions/points to make.
First, can we talk about how they simulated the âevolutionary modelâ in the paper? They simulated mutation and drift over 10,000 generations in a population of a constant size of 1000 individuals. This would be fine as a simple benchmark for comparison, but they repeatedly commented on how the results of this simulation diverged from the observed AFS, as though it was supposed to call into question the evolutionary model.
I kept waiting for them to comment on the obvious problem with this comparison, but it never came. The obvious fact is that no one is suggesting that the human population has been constant for the last 10,000 generations (never mind at 1000 individuals)! The âevolutionary modelâ includes a significant recent population growth, which can trivially be shown to shift the AFS to the left (more rare alleles) and produce a sharper bend between the far-left and middle-left of the AFS. Sanford et al. instead suggest that the evolutionary model must invoke a post hoc long-term bottleneck in the distant past. I suppose the more constant population size that existed for the many generations prior to the agricultural revolution and population explosion could be called a âbottleneckâ, but that doesnât really sound like what theyâre saying, and it certainly wouldnât be post hoc on our part.
Second, the âbiblical modelâ simulations include scenarios like starting with 2 individuals, letting the population grow for a few generations, then bottleneck to 6 individuals, and then grow for a couple of hundred generations. The final simulated population size is obviously very far from the current human population size - they donât include enough population growth. It seems to me that this would significantly change the AFS expected from their model, to produce a much steeper AFS (more rare variants). They discuss this briefly in the paper, in the part about âdemographic stirringâ, but I donât think theyâve adequately answered this question.
Third, while the AFS in terms of proportions of alleles with different frequencies in the population might be possible to get in their model of created heterozygosity, what about the actual number of different alleles? The more alleles you deem âoriginally createdâ, the smaller the proportion of âmutational allelesâ becomes, flattening out the AFS.
To tie this point into Nathaniel Jeansonâs work, consider the proportion of SNPs required to be âoriginally createdâ versus the proportion required to have arisen by mutation. Figures 6b and 7 of Sanford et alâs paper suggests that between 30-60% of all SNPs (discounting the rare alleles) in extant human populations would have been the result of mutations in the last few thousand years. On the other hand, in chapter 8 of his book âReplacing Darwinâ, Jeanson argues that âoriginally createdâ SNPs accounts for 99.6% of the variation in modern human populations, leaving just 0.4% of SNPs as the result of mutations (~15,000/4,000,000 - Figure 8.11 of the book). He does this because only around 15,000 mutations can separate the 2 most divergent living humans today given a 6,000-year timeframe, given current mutation rates.
I donât see how Jeansonâs claim can be reconciled with Sanford et alâs simulations of the AFS. I agree with Sanford et alâs logic in how the originally created SNPs would have to spread over the AFS from their original distributions at a frequency of 25% or 50% in Adam and Eve to form the low, flat distribution in the AFS, and I donât see how thatâs compatible with the vast majority of SNPs being pre-existing. Or am I mistaken in trying to compare these 2 things, because one is the AFS across a whole population while the other is the proportion of variants separating 2 individuals in a population?
Finally, the Y and mitochondrial chromosome AFS. Sanford et al. note that the actual AFS for these are distinct from the autosomes (figures 1b and 1c), and say that the AFS indicate these chromosomes are âyoungâ. I suppose thatâs true, relatively speaking. The Y chromosome and mtDNA have a more recent common ancestor than the entire autosomes do. However, can their âbiblical modelâ account for the alleles in a medium frequency in these chromosomes? For example, in their figure 1b, the bars between 15 and 30 on the x-axis? How about the actual number of SNPs? Jeanson shows in his book (based on his âpapersâ in ARJ) that the variation in the mtDNA can be (if he assumes a mitochondrial mutation rate thatâs about 4x higher than observed, mind you), but what about the Y chromosome? Surely there are more SNPs on the Y chromosome in modern human populations than could be explained in just 6,000 years, given current mutation rates?