Tuesday, June 03, 2008

Modeling human demographic history   posted by p-ter @ 6/03/2008 05:08:00 AM

In a previous post on current views on the human colonization of the world, I alluded two issues: whether modern humans displaced all archaics, and the precise demographic models under which that occurred. I placed more emphasis on the first, but was taken to task in the comments--apparently no one has issues with some version of the out-of-Africa story (ie. archaics contributed little genetic material to modern humans, if any), but there are some issues with the demographic models.

One paper pointed to is the Lohmueller et al. study on the high frequency of nonsynonymous polymorphisms in Europeans as compared to African-Americans. The authors do simulations under a variety of demographic scenarios, but rely heavily on a demography in which Europeans experienced a mild long-term bottleneck starting ~8000 generations (~160K years) ago, while the African populations instead experienced population growth. For those keeping track at home, humans aren't estimated to have left Africa until about 2000-4000 generations (40-80K years) ago. So these models rely on two populations having different demographies during a period of time before they split, an impressive feat indeed. So alright, some demographic histories aren't exactly the most believable.

That said, all bottlenecks are not created equal. The fact remains that as you move away from Africa, each population contains a subset of the diversity of the ones the precede it--this is indicative of each population being founded by a subset of those preceding individuals. And allele frequency spectra, when measured in different populations (see the figure here), are inconsistent with constant population size in Europe and East Asia, and imply instead a bottlenck, stronger in E. Asia than in Europe, in recent demographic history. Any reasonable model has to take these observations into account.