Estimating relative population sizes from simulated data sets and the question of greater African effective size

Abstract

Previous genetic and craniometric studies have suggested greater genetic diversity and a larger effective size in Africa. Several demographic scenarios can explain a larger African effective size, and anthropological geneticists have attempted to obtain better estimates of relative population sizes among continental regions in the Old World. A least-squares approach of estimating relative population weights was developed by Relethford and Harpending ([1994] Am. J. Phys. Anthropol. 95:249-270), who applied it to craniometric and genetic data sets and concluded that the ratio of African, Asian, and European effective sizes was 3:1:1, respectively; another data set of short tandem repeat (STR) markers yielded a similar estimate of 7:1:2. However, an estimate from restriction site polymorphism (RSP, also known as restriction fragment length polymorphism, or RFLP) data yielded a very different estimate of 1:1:8. Thus, the European and not the African effective size was largest. Simulations showed that this was the result of ascertainment bias in which polymorphic markers were originally identified in a small panel of Caucasians, leading to inflated heterozygosity in the European sample and thus an inflated population weight. This paper extends those computer simulations to incorporate not only ascertainment bias but also interpopulation gene flow and demographic expansion in two types of genetic data, single nucleotide polymorphisms (SNPs, which are similar but not precisely identical to RSPs) and STRs. The effects of these three parameters on SNP and STR relative weight estimates are described. Simulations show that the ascertainment scheme affects SNP data but not STR data. Gene flow has a noticeable effect on the bias and efficiency of the estimates in both types of genetic data. Population expansions have a large effect only in one ascertainment scheme in the simulated SNP data and no effect in STR data. Relative population weight estimates from four published STR data sets are also reported. These estimates are similar to each other: all show a larger African weight and a European weight somewhat larger than the Asian weight. Because the STR simulations show that when gene flow is greater than 0.01 migrants per generation the African population weight is biased upward, it is likely that the African weights in the four STR data sets are inflated. However, the simulations suggest that the African effective size is still largest of the three regions and is probably at least as great as the sum of the Asian and European effective sizes.