Of course, the average amount of genetic variation is simply a per locus average. This hides two important dimensions of the problem: first, that genes don't have independent effects on phenotypes, which often have a polygenic mode of inheritance. If a number of different genes influence a particular phenotype, then small average differences in each of them may amount to a large overall difference.
But, the fact that the average amount of ~15% is average means that individual loci themselves may exhibit a much higher degree of genetic differentiation, something which is not immediately apparent by considering the average.
A new genetic study has found one such locus in the CYP3A5 gene which affects "salt and water retention and risk for salt-sensitive hypertension":
Whereas average FST values in worldwide human populations are ∼0.123 (Akey et al. 2002), seven and two SNPs in CYP3A4 and CYP3A5, respectively, had an FST >0.5 between African Americans and the non-Africans, reaching a maximum of 0.66.
Note that unlike many ancestry informative polymorphisms which are defined in so-called "junk" DNA, such as e.g., markers defining Y-chromosomal or mtDNA haplogroups, the CYP3A5 locus has a functional importance, and as the authors observe, has undergone selection in non-African population, leading to the current pattern of strong differentiation.
Am. J. Hum. Genet., 75:000, 2004 (Early View)
CYP3A Variation and the Evolution of Salt-Sensitivity Variants
E. E. Thompson et al.
Members of the cytochrome P450 3A subfamily catalyze the metabolism of endogenous substrates, environmental carcinogens, and clinically important exogenous compounds, such as prescription drugs and therapeutic agents. In particular, the CYP3A4 and CYP3A5 genes play an especially important role in pharmacogenetics, since they metabolize >50% of the drugs on the market. However, known genetic variants at these two loci are not sufficient to account for the observed phenotypic variability in drug response. We used a comparative genomics approach to identify conserved coding and noncoding regions at these genes and resequenced them in three ethnically diverse human populations. We show that remarkable interpopulation differences exist with regard to frequency spectrum and haplotype structure. The non-African samples are characterized by a marked excess of rare variants and the presence of a homogeneous group of long-range haplotypes at high frequency. The CYP3A5*1/*3 polymorphism, which is likely to influence salt and water retention and risk for salt-sensitive hypertension, was genotyped in >1,000 individuals from 52 worldwide population samples. The results reveal an unusual geographic pattern whereby the CYP3A5*3 frequency shows extreme variation across human populations and is significantly correlated with distance from the equator. Furthermore, we show that an unlinked variant, AGT M235T, previously implicated in hypertension and pre-eclampsia, exhibits a similar geographic distribution and is significantly correlated in frequency with CYP3A5*1/*3. Taken together, these results suggest that variants that influence salt homeostasis were the targets of a shared selective pressure that resulted from an environmental variable correlated with latitude.