Founding of Eurasian population: 52ky (?)

The authors write:

As in GUTENKUNST et al. (2009), we assumed a divergence time of 6 million years between human and chimpanzee and a generation time of 25 years. We estimated a mutation rate of μ = 2.35 × 10−8 per site per generation from sequence divergence present in the data. This is identical to the estimate by GUTENKUNST et al. (2009) and comparable to other estimates (e.g. NACHMAN and CROWELL (2000)).

Nachman and Crowell's rate is derived under an assumption of the human-chimp divergence, and of the human ancestral population size. Gutenkurst et al. inferred a 100,000-year old divergence time with the same mutation rate as Lukic and Hey. But, direct measurements of the human autosomal mutation rate range between 1-1.3x10^-8, with a recent estimate of 1.38x10^-8 from a large population sample of ~14,000 individuals where Ne and μ could be simultaneously estimated.

As I mentioned before, the discrepancy between the low and high autosomal mutation rate is substantial. But, we must remember that the widely-cited 2.5x10^-8 mutation rate of Nachman and Crowell depends on some very strict assumptions: 5 million years between chimp and human, 20-year generation: change the effective size to 10^5 individuals, and the mutation rate drops to 1.5x10^-8.

This is an important distinction to resolve: if the slower rate is adopted, then Out-of-Africa may go to >100ka, and may correspond to the Nubian Complex of Arabia. With the faster rate, we get the archaeologically invisible 60,000 year old coastal migration. I tend to favor the older time, but the issue needs to be addressed.

Genetics August 3, 2012 genetics.112.141846

Demographic Inference Using Spectral Methods on SNP Data, With an Analysis of the Human out-of-Africa Expansion

Sergio Lukić, and Jody Hey

We present an implementation of a recently introduced method for estimating the allele frequency spectrum under the diffusion approximation. For SNP frequency data from multiple populations, the method computes numerical solutions to the allele frequency spectrum (AFS) under a complex model that includes population splitting events, migration, population expansion and admixture. The solution to the diffusion partial differential equation (PDE) that mimics the evolutionary process is found by means of truncated polynomial expansions. In the absence of gene-flow, our computation of frequency spectra yields exact results. The results are compared to those which use a finite difference method and to forward diffusion simulations. In general, all the methods yield comparable results, although the polynomial-based approach is the most accurate in the weak-migration limit. Also, the economical use of memory attained by the polynomial expansions makes the study of models with four populations possible for the first time. The method was applied to a four-population model of the human expansion out of Africa and the peopling of the Americas, using the Environmental Genome Project (EGP) SNP database. Although our confidence intervals largely overlapped previous analyses of these data, some were significantly different. In particular, estimates of migration among African, European and Asian populations were considerably lower than in a previous study and the estimated time of migration out of Africa was earlier. The estimated time of founding of a human population outside of Africa was 52,000 years (95% confidence interval: 36,000 - 80,800 years).

Link