The de-brownification of brown people?

Share on FacebookShare on Google+Email this to someoneTweet about this on Twitter

About a year ago a paper came out, Low Levels of Genetic Divergence across Geographically and Linguistically Diverse Populations from India. The authors used Asian Indian groups from the United States, ergo, the caste/class representativeness is not is very typical. Additionally, there is a strong skew to Gujaratis since this group represents 1/2 of American Asian Indians. One could offer the reasonable opinion that the low amount of between population variance is simply a function of the fact that higher status groups across South Asia are not particularly differentiated from each other. I don’t have a final opinion, but I would be moderately skeptical of this because I’ve seen enough work in the past suggesting that Brahmins, for example, are not particularly closely related to each across regions (and many of these regional Brahmin groups show strong evidence of gene flow with other local caste groups, though there usually there is some differentiation).

In any case, that’s just a preface to the fact that a provisional paper has come out from the same group, Prevalence of common disease-associated variants in Asian Indians. If you’re interested in the topic of the paper, all I have to say is that it seems that the major inference one might make is that more studies need to be done with Asian Indians because they likely have a whole lot of population-specific disease variants which aren’t well known yet. Human biodiversity has some practical medical implications. But that’s not what I want to focus on. In Table 4 they have some data on frequency of the minor frequency allele on SLC24A5, which in the case of South Asians is the ancestral variant. That is, the allele which is fixed in Africans and East Asians, and absent in Europeans, is the one shown in the table.

Language N Frequency of ancestral allele
Assamese 26 0.260
Bengali 27 0.293
Gujarati 181 0.208
Hindi 29 0.105
Kannada 24 0.268
Kashmiri 24 0.020
Konkani 43 0.058
Malayalam 25 0.259
Marathi 26 0.296
Marwari 25 0.041
Oriya 27 0.154
Parsi 25 0.100
Punjabi 28 0.052
Tamil 29 0.036
Telugu 28 0.000

There’s nothing that surprising here. I wouldn’t take some of the frequencies as scripture; I suspect that some of these linguistic groups are not representative (the frequency for the Gujaratis I can believe since I’ve seen it elsewhere, while the northwestern groups are in line with the ones from the HGDP populations). It seems likely that a strong NW-SE cline on the variation of this gene exists just as it does on many genes in South Asia. There is data which shows that in Sri Lanka the frequency of the ancestral allele is 0.75 for Tamils, and 0.50 for Sinhalese. I wouldn’t read too much into these data either; but I’ve seen results elsewhere which would imply that you shouldn’t be surprised at seeing a ancestral allele frequency around ~0.25 for South Asians on SLC24A5. I know this is a provisional paper, and this might get yanked, but I want to point this out:

The MAF [minor allele frequency] for the SLC24A5 g.13242G>A polymorphism [the ancestral allele] (0.114) in the Indian population is closer to that of Caucasian European populations (0.000-0.020) than to that of East Asians (0.979-0.989) and Africans (0.730-0.980)…This is surprising given that, as a whole, people of Indian origin have darker skin tone compared to Europeans. The East Asian and African populations share a similarly high frequency of the minor allele of the SLC24A5 g.13242G>A SNP, which in itself is surprising as East Asians are typically of much lighter skin than Africans. It is also interesting that the t-test suggested a greater similarity in MAF between Indians and those of populations in the Americas (Puerto Rican, Mexican, and Amerindian), who are of a similar skin tone….

No shit it’s surprising. We know that SLC24A5 polymorphism can account for about 1/3 of the skin color variation in South Asians. It isn’t as if it doesn’t affect skin color in this population due to some modifier locus. The frequencies of SLC24A5 alleles in Latin American populations is just a proxy for European and non-European admixture; the non-European groups in Latin America are Amerindians and Africans, both of whom tend to carry the ancestral variant.

In any case, contrary to the impressions of those of you who only know of South Asians from Bollywood, Indians are on average darker-skinned than Puerto Ricans or Mestizos. SLC24A5 has been under selection within the last 10,000 years. It explains 25-40% of the between group difference in complexion of Europeans and Africans (checked via an admixture study with African Americans). But there’s no way I think that light skin was being selected in southern India within the last 10,000 years. Something else is going on….



  1. It would have been better had they included SLC45A2 gene in their study.

  2. for 45a2, almost zero in southern india. on the order of 0.10 to 0.25 in northern india. for the derived “light skin” allele.

  3. One interesting observation – despite Malayalees and Tamilians being “lingustic brothers in arms”, they have an order of magnitude delta in the allele (0.296 vs. 0.036).  
    Any guesses how this comes about?

  4. Thanks, Razib. I am rather fascinated by Soejima et al. explanation on that allele. 
    There are several scenarios to 
    explain the differences in allele frequency of the coding 
    SNPs between the SLC24A5 and SLC45A2 genes in South 
    Asian populations. One of them is that 374L has a selective 
    advantage against harmful ultraviolet radiation (UVB), and 
    the frequency of the 374F allele migrating from Central 
    Asian populations might have been reduced by purifying 
    selection in South Asian populations. This may mean that 
    the functional constraint is stronger on SLC45A2 than on 
    SLC24A5. Sequence analysis of the SLC24A5 gene flanking 
    the p.A111T polymorphism and functional analysis of the 
    coding SNPs will help us understand the evolution of these 
    two pigmentation genes.
    Its frequency among Brahmin dominated south Indian population in the US (Tamils and Kannadigas…not so sure about Telugus and Malayalis) would be interesting to know.

  5. v-man, representativeness. american gujaratis are mostly patels. malayalees are disproportionately xtian. tamils? i assume way overloaded with brahmins. you might be seeing distortions because of population substructure within tamil nadu. that’s why i brought up sri lanka. 
    manju, there might be purifying selection. but there is no way that SLC24A5 frequencies in south asia are due simply to admixture. implies WAY too much exogenous input in the last 10,000 years.

  6. Sorry to bring unfounded knowledge into this, but aren’t there stories that hinduism was established by light skinned Persians? If the religious leaders were light skinned then clearly there is an adaptive advantage to being light skinned.

  7. Sorry to bring unfounded knowledge into this, but aren’t there stories that hinduism was established by light skinned Persians? If the religious leaders were light skinned then clearly there is an adaptive advantage to being light skinned. 
    #1, not really. the vedic component of hinduism is indo-aryan (not iranian though), but i think most religious scholars would make the case that hinduism is less indo-aryan than christianity is jewish. IOW, the “indigenous” religious ideas are much more prominent and central to classical and modern hinduism than the indo-aryan ones. “reformist” groups like arya samaj go back to the vedas…but they also introduce some real modern ideas. 
    as for #2, i think there is something to this. but the selection coefficients are too big; it isn’t like priests were 30% of the population. additionally, south india wasn’t hindu until the last 2,000 years. IOW, the numbers don’t work out, though it could be some of the effect….