I agree, although background selection in particular is very weak given human recombination rates and gene density, so that would take a lot more selection (across deleterious variants), while hitchhiking would take relatively few instances of positive selection.

Selection on standing variants may be a factor, although (a) they are rare relative to to new mutations, because the human population formerly was very small, and (b) if they’re old enough, they shouldn’t be tightly linked to particular allelic variants, even where recombination is low.

I’ll do updated numbers on both these factors and post them, because it’s important to get those weights.

Which is exactly what you’d expect if there were a lot of new selected alleles at frequencies less than 20 percent that had pulled common variants along with them, over a longer region in areas of lower recombination.

This is key to the literature on “draft”. The usual observation is that variation is correlated with recombination rate, because of the effect of hitchhiking on common linked neutral alleles.

So in light of that, let me give a different example (I admit I sort of randomly pulled those initial numbers out of the air without much thought):

Allele 1: 60% in Japan, 50% in China
Allele 2: 55% in Japan, 50% in China

The observation (again, I think) in this paper is that the former are more common in regions of low recombination compared to the latter.

Yep, that’s what I mean.

Right. So we agree, it has nothing to do with low frequency, geographically restricted alleles :)

mean Fst between china and japan in the hapmap is ~0.005, fwiw.

It may help you to try running some numbers. Your example — 55% in Japan, 50% in China — leads to an Fst of 0.002, which is of course quite a bit less than the mean Fst we see between Japan and China. To get a substantial increase over the mean Fst (which is around 0.02-0.04 within continents), you would need allele frequencies that differ like more than 0.5 and 0.7 or so.

Keeping those numbers in mind, you can see that it actually takes a fairly substantial shift in frequencies in a low-recombination block to make two nearby populations look more divergent for that particular genomic region compared to the mean.

The most plausible way to get that kind of shift in low-frequency regions is hitchhiking.

It’s a little unfortunate that they don’t report the genic/nongenic comparison for the HapMap comparisons, but that will be easy enough to check.