Cited by 161melanin and skin pigmentationForensic and Genetic ResearchGenetic Associations and Epidemiology

The DOI matches its Crossref record; 54 of the 55 references it cites were found in a registry or an index; 3 of the 39 sources its claims cite were checked for support.

Genetic Architecture of Skin and Eye Color in an African-European Admixed Population

Open access

Abstract

Variation in human skin and eye color is substantial and especially apparent in admixed populations, yet the underlying genetic architecture is poorly understood because most genome-wide studies are based on individuals of European ancestry. We study pigmentary variation in 699 individuals from Cape Verde, where extensive West African/European admixture has given rise to a broad range in trait values and genomic ancestry proportions. We develop and apply a new approach for measuring eye color, and identify two major loci (HERC2[OCA2] P = 2.3 × 10(-62), SLC24A5 P = 9.6 × 10(-9)) that account for both blue versus brown eye color and varying intensities of brown eye color. We identify four major loci (SLC24A5 P = 5.4 × 10(-27), TYR P = 1.1 × 10(-9), APBA2[OCA2] P = 1.5 × 10(-8), SLC45A2 P = 6 × 10(-9)) for skin color that together account for 35% of the total variance, but the genetic component with the largest effect (~44%) is average genomic ancestry. Our results suggest that adjacent cis-acting regulatory loci for OCA2 explain the relationship between skin and eye color, and point to an underlying genetic architecture in which several genes of moderate effect act together with many genes of small effect to explain ~70% of the estimated heritability.

Overview

AI snapshot

Written from the paper’s full text.

TL;DR

This study maps skin and eye color in 699 Cape Verdeans, an African-European admixed population, using genotype- and ancestry-based association approaches. It finds distinct genetic architectures: a few major loci plus average genomic ancestry explain much of skin color variation, while eye color is dominated by HERC2[OCA2] with a secondary SLC24A5 signal.

Key findings

  • Two major eye-color loci were identified: HERC2[OCA2] rs12913832 (P = 2.3×10−62) and SLC24A5 (P = 9.6×10−9 in the abstract; Table 1 reports rs2470102 genotype P = 4.8×10−10), affecting both blue versus brown eyes and shades of brown.
  • Four major skin-color loci were identified: SLC24A5 (P = 5.4×10−27), TYR/GRM5 (P = 1.1×10−9), APBA2[OCA2] (P = 1.5×10−8), and SLC45A2 (P = 6×10−9); together they account for 35% of total skin-color variance, while average genomic ancestry accounts for ∼44%.
  • Skin and eye color have different architectures: skin color correlates with African genomic ancestry at R² = 0.44, whereas eye color correlates weakly with ancestry (R² = 0.08) and with skin color (R² = 0.14).
  • At 15q13.1, the eye-color signal in HERC2[OCA2] and the skin-color signal in APBA2[OCA2] are genetically separable, with weak linkage disequilibrium between peaks (R² = 0.112), suggesting adjacent cis-regulatory effects on OCA2.

Methods

  • Studied 699 Cape Verdeans from six islands; final analyses used 685 individuals with skin reflectance and 625 with high-quality eye photographs. Genotyping used Illumina 1M, yielding 903,837 autosomal SNPs after QC.
  • Skin color was measured as a modified melanin (MM) index by reflectance spectroscopy on the upper inner arm; eye color was measured as a new quantitative “T index” derived from digital photographs and a principal curve.
  • Association analyses used additive linear regression adjusted for sex and first three principal components, plus ancestry-based mapping with SABER+, EMMAX mixed models, imputation with MaCH, CMS selection scores, and power simulations.

Limitations

  • The paper states that CEU and YRI are imperfect proxies for the true ancestral populations of Cape Verde, which likely include multiple Southern European and West African groups, and cautions that “European” and “African” should not imply homogeneity.
  • The authors note that GCTA-style heritability estimation was not used because the genetic relationship matrix is inflated by population structure and pilot studies showed large sampling variability in this moderate-sized sample.
  • The paper states that failure to replicate some previously reported loci (e.g., ASIP, KITLG, MC1R) could reflect limited ancestral allelic variation, incomplete ancestry correction in earlier studies, or effects too small to detect; the APBA2[OCA2] regulatory mechanism remains a hypothesis requiring functional validation.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1371/journal.pgen.1003372

Cites

54 of 54 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Relationship of geography and ancestry to skin and eye color. Individual ancestry proportions for Cape Verdeans displayed on all four panels were obtained from a supervised analysis in frappe with K = 2 and HapMap's CEU and YRI fixed as European and African parental populations. (a) Bar plots of individual ancestry proportions for Cape Verdeans across the islands. The width of the plots is proportional to sample size (Santiago, n = 172; Fogo, n = 129; NW cluster, n = 192; Boa Vista, n = 27). The proportion of African vs. European ancestry of the individuals is indicated by the proportion of blue vs. red color in each plot. (b) Individual African ancestry distribution in the total cohort of 685 Cape Verdeans (histogram) and in 802 African Americans (kernel density curve) from the Family Blood Pressure Program (FBPP) [21]. (c) Scatter-plot of skin color vs. Individual African ancestry proportions. Skin color is measured by the MM index described in Material and Methods. (d) Scatter-plot of eye color vs. Individual African ancestry proportions. Eye color is measured by the T-index, described in Figure 2 and Material and Methods. Points in scatter-plots are color coded according to the island of origin of the individuals.
Figure 1 Relationship of geography and ancestry to skin and eye color. Individual ancestry proportions for Cape Verdeans displayed on all four panels were obtained from a supervised analysis in frappe with K = 2 and HapMap's CEU and YRI fixed as European and African parental populations. (a) Bar plots of individual ancestry proportions for Cape Verdeans across the islands. The width of the plots is proportional to sample size (Santiago, n = 172; Fogo, n = 129; NW cluster, n = 192; Boa Vista, n = 27). The proportion of African vs. European ancestry of the individuals is indicated by the proportion of blue vs. red color in each plot. (b) Individual African ancestry distribution in the total cohort of 685 Cape Verdeans (histogram) and in 802 African Americans (kernel density curve) from the Family Blood Pressure Program (FBPP) [21]. (c) Scatter-plot of skin color vs. Individual African ancestry proportions. Skin color is measured by the MM index described in Material and Methods. (d) Scatter-plot of eye color vs. Individual African ancestry proportions. Eye color is measured by the T-index, described in Figure 2 and Material and Methods. Points in scatter-plots are color coded according to the island of origin of the individuals.CC BY · PubMed Central
Figure 2 Quantitative assessment of eye color. Plotted are the normalized median values of green (x-axis) and blue (y-axis) levels of each individual's irises. We fitted a principal curve that explains most of the variation in the data (red dashed curve). The T-index is defined by the arc-length from the projection of each point on the curve to the end of the curve that corresponds to the lightest eye color. In the figure are examples of eye photos at their respective position in the T-index curve.
Figure 2 Quantitative assessment of eye color. Plotted are the normalized median values of green (x-axis) and blue (y-axis) levels of each individual's irises. We fitted a principal curve that explains most of the variation in the data (red dashed curve). The T-index is defined by the arc-length from the projection of each point on the curve to the end of the curve that corresponds to the lightest eye color. In the figure are examples of eye photos at their respective position in the T-index curve.CC BY · PubMed Central
Figure 3 GWAS results for skin and eye color in the total Cape Verdean cohort. Results are shown as −log10(P value) for the genotyped SNPs. Plots are ordered by chromosomal position. (a,c) Genotype and admixture association scan results for skin color. (b,d) Genotype and admixture association scan results for eye color. (a,b) show the P values obtained in the initial scans and (c,d) the P values of the following scans adjusting for the strongest associated SNP (in SLC24A5 for skin color and in HERC2 for eye color). Dashed red lines correspond to the genome-wide significance threshold (P<5×10−8 in the genotype scan; P<7×10−6 in the ancestry scan [see Material and Methods]). The location and identity of candidate genes are colored to correspond with chromosomal location; individual SNPs are given in Table 1.
Figure 3 GWAS results for skin and eye color in the total Cape Verdean cohort. Results are shown as −log10(P value) for the genotyped SNPs. Plots are ordered by chromosomal position. (a,c) Genotype and admixture association scan results for skin color. (b,d) Genotype and admixture association scan results for eye color. (a,b) show the P values obtained in the initial scans and (c,d) the P values of the following scans adjusting for the strongest associated SNP (in SLC24A5 for skin color and in HERC2 for eye color). Dashed red lines correspond to the genome-wide significance threshold (P<5×10−8 in the genotype scan; P<7×10−6 in the ancestry scan [see Material and Methods]). The location and identity of candidate genes are colored to correspond with chromosomal location; individual SNPs are given in Table 1.CC BY · PubMed Central
Figure 4 Imputation, fine-mapping, and selective signature scores for skin and eye color loci. Each panel shows the genotype-association results and the distribution of Composite of Multiple Signals scores of positive selection (CMS scores) [38] for genotyped and imputed SNPs (dark and light blue points in the association plots, respectively) surrounding the candidate loci; identities of individual SNPs are given in Table 1. Dashed red lines correspond to the genome-wide significance thresholds. (a) 5p13.3 region containing the SLC45A2 gene. (b) 11q14.3 region containing the GRM5 and TYR genes. (c) 15q21.1 region containing the SLC24A5 gene. (d) 15q13.1 region containing the OCA2, HERC2 and APBA2 genes. The interval between HERC2 and APBA2 contains a cluster of segmental duplications and a paucity of SNPs [45], and is also a source of a frequent deletion breakpoint as described in the Discussion. Previously identified candidate causative non-synonymous variants [11], [13], [14] are denoted as red dots in the association plots of panels (a,b,c). The location and identity of candidate genes are colored to correspond with chromosomal location; individual SNPs are given in Table 1.
Figure 4 Imputation, fine-mapping, and selective signature scores for skin and eye color loci. Each panel shows the genotype-association results and the distribution of Composite of Multiple Signals scores of positive selection (CMS scores) [38] for genotyped and imputed SNPs (dark and light blue points in the association plots, respectively) surrounding the candidate loci; identities of individual SNPs are given in Table 1. Dashed red lines correspond to the genome-wide significance thresholds. (a) 5p13.3 region containing the SLC45A2 gene. (b) 11q14.3 region containing the GRM5 and TYR genes. (c) 15q21.1 region containing the SLC24A5 gene. (d) 15q13.1 region containing the OCA2, HERC2 and APBA2 genes. The interval between HERC2 and APBA2 contains a cluster of segmental duplications and a paucity of SNPs [45], and is also a source of a frequent deletion breakpoint as described in the Discussion. Previously identified candidate causative non-synonymous variants [11], [13], [14] are denoted as red dots in the association plots of panels (a,b,c). The location and identity of candidate genes are colored to correspond with chromosomal location; individual SNPs are given in Table 1.CC BY · PubMed Central
Figure 5 Power and candidate gene analyses. (a) Power estimated at three different alpha levels, plotted as a function of effect size (number of MM units by which each copy of the derived allele lightens skin pigmentation). The results shown here are based on derived allele frequencies of 0.1 and 0.9 in the ancestral African and European populations, respectively. The effect sizes detected in Cape Verde for the four major skin color loci are shown in blue; effect sizes for ASIP and KITLG as estimated (see Material and Methods) from [12] and [15], respectively are shown in red. (b) Distribution of P values for SNPs in 47 candidate genes, 16 regions with strong signatures of selection, and random SNPs, all shown as a q-q plot of the −log10 (P) values. Observed −log10 (P) values specifically for ASIP and KITLG SNPs are shown above the plot.
Figure 5 Power and candidate gene analyses. (a) Power estimated at three different alpha levels, plotted as a function of effect size (number of MM units by which each copy of the derived allele lightens skin pigmentation). The results shown here are based on derived allele frequencies of 0.1 and 0.9 in the ancestral African and European populations, respectively. The effect sizes detected in Cape Verde for the four major skin color loci are shown in blue; effect sizes for ASIP and KITLG as estimated (see Material and Methods) from [12] and [15], respectively are shown in red. (b) Distribution of P values for SNPs in 47 candidate genes, 16 regions with strong signatures of selection, and random SNPs, all shown as a q-q plot of the −log10 (P) values. Observed −log10 (P) values specifically for ASIP and KITLG SNPs are shown above the plot.CC BY · PubMed Central
Figure 6 Allele frequency and haplotype analysis for eye and skin color loci at 15q13.1. (a,b) Allele frequency distributions for the SNPs most significantly associated with eye [HERC2 rs1291382; (a)] and skin [APBA2 rs4424881; (b)] in the HapmapIII [19] and HGDP [39] panels for the old world. Blue/orange shading corresponds to the frequency of the ancestral/derived alleles, as determined by comparison with the chimp reference sequence (assembly CGSC2.1/pan Tro2). Frequency values are presented in Table S3. c) Visual displays of the haplotypes extending from HERC2 to the second intron of APBA2 in Europeans from HapMap phase III (CEU) and HGDP (French, French Basque, North Italian, Tuscan, Sardinian, Orcadian, and Russian) panels. Haplotypes were inferred on the basis of 26 SNPs common to both datasets; blue and orange shades represent the ancestral and derived alleles, respectively. Haplotypes were ordered according to the ancestral/derived states at HERC2 rs1291382 and APBA2 rs4424881 (marked with red arrows), as follows: haplotypes bearing the ancestral alleles for both SNPs (Anc-Anc); haplotypes bearing the derived allele for HERC2 rs1291382 and the ancestral allele for APBA2 rs4424881 (Der-Anc); haplotypes bearing the ancestral allele for HERC2 rs1291382 and the derived allele for APBA2 rs4424881 (Anc-Der); and haplotypes bearing the derived allele for both SNPs (Der-Der).
Figure 6 Allele frequency and haplotype analysis for eye and skin color loci at 15q13.1. (a,b) Allele frequency distributions for the SNPs most significantly associated with eye [HERC2 rs1291382; (a)] and skin [APBA2 rs4424881; (b)] in the HapmapIII [19] and HGDP [39] panels for the old world. Blue/orange shading corresponds to the frequency of the ancestral/derived alleles, as determined by comparison with the chimp reference sequence (assembly CGSC2.1/pan Tro2). Frequency values are presented in Table S3. c) Visual displays of the haplotypes extending from HERC2 to the second intron of APBA2 in Europeans from HapMap phase III (CEU) and HGDP (French, French Basque, North Italian, Tuscan, Sardinian, Orcadian, and Russian) panels. Haplotypes were inferred on the basis of 26 SNPs common to both datasets; blue and orange shades represent the ancestral and derived alleles, respectively. Haplotypes were ordered according to the ancestral/derived states at HERC2 rs1291382 and APBA2 rs4424881 (marked with red arrows), as follows: haplotypes bearing the ancestral alleles for both SNPs (Anc-Anc); haplotypes bearing the derived allele for HERC2 rs1291382 and the ancestral allele for APBA2 rs4424881 (Der-Anc); haplotypes bearing the ancestral allele for HERC2 rs1291382 and the derived allele for APBA2 rs4424881 (Anc-Der); and haplotypes bearing the derived allele for both SNPs (Der-Der).CC BY · PubMed Central
Figure 7 Genetic architecture of skin color variation. (a) Effect sizes of the loci associated with skin color. Effect values represent the beta values obtained from a regression model containing the four associated loci plus ancestry. (b) The pie chart represents the proportion of phenotypic variation accounted for by the different components, including non-heritable factors (∼20%), the four major loci (∼35%, color-coded as in [a]), and average genomic ancestry (44%). The heritable contributions were estimated by regression and variance decomposition as described in Material and Methods, and are also represented below the pie chart separately as grey (genomic ancestry) or open (four major loci) areas. However, because of admixture stratification, the heritable contributions overlap as described in the text.
Figure 7 Genetic architecture of skin color variation. (a) Effect sizes of the loci associated with skin color. Effect values represent the beta values obtained from a regression model containing the four associated loci plus ancestry. (b) The pie chart represents the proportion of phenotypic variation accounted for by the different components, including non-heritable factors (∼20%), the four major loci (∼35%, color-coded as in [a]), and average genomic ancestry (44%). The heritable contributions were estimated by regression and variance decomposition as described in Material and Methods, and are also represented below the pie chart separately as grey (genomic ancestry) or open (four major loci) areas. However, because of admixture stratification, the heritable contributions overlap as described in the text.CC BY · PubMed Central

Verification

Claims the paper makes

56 Own result · 33 Not checked yet · 30 Own interpretation · 18 Own method · 9 No basis given · 1 Partly supported · accuracy not yet measured

Verify all 36 references — free account

Show all 75 claims

  • It may state two separate claims as one. 58
  • The quote only partly states this claim. 1
  • Probably not a substantial claim. 23
72 of 72
  • The paper relies on this without support.

    The true ancestral populations for Cape Verde likely include contributions from several areas of Southern Europe and several areas of West Africa.

    “We also note that the true ancestral populations for Cape Verde are likely to include contributions from several areas of Southern Europe, and several areas of West Africa.”
  • The authors state this as an assumption.

    Skin color in Cape Verde is 80% heritable.

    “If skin color in Cape Verde is 80% heritable, the four major loci and individual ancestry explain most of that variation”
  • The paper relies on this without support.

    HapMap CEU and YRI individuals adequately represent the European and African ancestral populations of Cape Verde.

    “Data from CEU and YRI HapMap individuals [19] was used to partition each Cape Verdean individual's ancestry into “European” and “African” components.”
  • The authors state this as an assumption.

    The Cape Verdean population has two ancestral populations (European and West African).

    “assuming two ancestral populations”
  • The authors state this as an assumption.

    Residual non-genetic variance is independent of any genotype or ancestry.

    “The residual non-genetic variance was assumed to be independent of any genotype or ancestry, such that the total variance in each simulated phenotype matches that observed for skin pigmentation in Cape Verde.”
  • The authors state this as an assumption.

    The simulated phenotype has a genetic architecture with three major loci, each with the same effect as SLC24A5 and perfectly ancestrally informative.

    “The phenotypes were simulated assuming a genetic architecture with three major loci, each of which has the same effect as SLC24A5 and is perfectly ancestrally informative.”
6 of 6

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

2 of 2

Citations

Cites

Show all 55 references

Research works printed but not linked: 1

  1. Baleno IC (2001) Povoamento e Formação da Sociedade. In: Albuquerque L, Santos, M.E., editor. História Geral de Cabo Verde. Lisbon, Praia: Instituto de Investigação Científica Tropical, Instituto Nacional de Investigação, Promoção e Património Culturais de Cabo Verde. pp. 125–177.

Cited by

Sorted by citations

Metadata

Authors
Beleza, S., Johnson, N. A., Candille, S. I., Absher, D. M., Coram, M. A., Lopes, J., Campos, J., Araújo, I. I., Anderson, T. M., Vilhjálmsson, B. J., Nordborg, M., Correia e Silva, A., Shriver, M. D., Rocha, J., Barsh, G. S., & Tang, H.
Affiliations
  • Universidade do Porto (PT)
  • Stanford University (US)
  • Stanford Medicine (US)
  • HudsonAlpha Institute for Biotechnology (US)
  • University of Cape Verde (CV)
  • Gregor Mendel Institute of Molecular Plant Biology (AT)
  • Pennsylvania State University (US)
Corresponding author
  • Beleza, Sandra
  • Barsh, Gregory S.
Venue
PLoS Genetics
Published
2013-03-21
Volume
9
Issue
3
Pages
e1003372
ISSN
  • 1553-7390 (ISSN-L)
  • 1553-7404
OpenAlex
W2129233990
Cited by
161
References
55
Citations by year
2026: 6 · 2025: 7 · 2024: 8 · 2023: 10 · 2022: 12 · 2021: 20 · 2020: 11 · 2019: 9 · 2018: 13 · 2017: 20 · 2016: 11 · 2015: 15 · 2014: 12 · 2013: 7
FWCI
8.70
Type
journal article
Language
en
Open access
Locations
  • PLoS Genetics (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • Figshare (submittedVersion, cc-by)
Funders
  • Fundação para a Ciência e a Tecnologia
  • National Human Genome Research Institute (T32HG000044)
  • National Institute of General Medical Sciences (GM073059, T32GM007790, R01GM073059)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Cell Biology › melanin and skin pigmentation
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Forensic and Genetic Research
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genetic Associations and Epidemiology
Keywords
skin color, eye color, admixed populations, Cape Verde, SLC24A5, OCA2, HERC2, TYR, SLC45A2, genetic ancestry, heritability
MeSH terms
Eye Color, Genotype, Hair Color, Haplotypes, Humans, Skin Pigmentation, Albinism, Oculocutaneous, Polymorphism, Single Nucleotide, Black People, White People, Cabo Verde

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review