Genome-wide patterns of population structure and
admixture in West Africans and African Americans
Katarzyna Bryca, Adam Autona, Matthew R. Nelsonb, Jorge R. Oksenbergc, Stephen L. Hauserc, Scott Williamsd,
Alain Fromente, Jean-Marie Bodof, Charles Wambebeg, Sarah A. Tishkoffh,1,2, and Carlos D. Bustamantea,1,3
aDepartment of Biological Statistics and Computational Biology, Cornell University, Ithaca, NY 14853;bGlaxoSmithKline, Research Triangle Park, NC 27709;
cDepartment of Neurology, University of California, San Francisco, CA 94143;dDepartment of Molecular Physiology and Biophysics, Center for Human
Genetics Research, Vanderbilt University, Nashville, TN 37232;eUnité Mixte de Recherche 208, Institut de recherche pour le développement (IRD)-Muséum
national d’Histoire naturelle (MNHN), Musée de l’Homme, 75116 Paris, France;fMinistére de la Recherche Scientifique et de l’Innovation, BP 1457, Yaoundé,
Cameroon;gInternational Biomedical Research in Africa, Abuja, Nigeria; andhDepartments of Genetics and Biology, University of Pennsylvania, Philadelphia,
Edited by Mary-Claire King, University of Washington, Seattle, WA, and approved November 19, 2009 (received for review August 25, 2009)
Quantifying patterns of population structure in Africans and
African Americans illuminates the history of human populations
and is critical for undertaking medical genomic studies on a global
scale. To obtain a fine-scale genome-wide perspective of ancestry,
we analyze Affymetrix GeneChip 500K genotype data from
African Americans (n = 365) and individuals with ancestry from
West Africa (n = 203 from 12 populations) and Europe (n = 400 from
42 countries). We find that population structure within the West
African sample reflects primarily language and secondarily geo-
graphical distance, echoing the Bantu expansion. Among African
Americans, analysis of genomic admixture by a principal compo-
nent-based approach indicates that the median proportion of Euro-
pean ancestry is 18.5% (25th–75th percentiles: 11.6–27.7%), with
very large variation among individuals. In the African-American
sample as a whole, few autosomal regions showed exceptionally
high or low mean African ancestry, but the X chromosome showed
elevated levels of African ancestry, consistent with a sex-biased
pattern of gene flow with an excess of European male and African
female ancestry. We also find that genomic profiles of individual
African Americans afford personalized ancestry reconstructions
differentiating ancient vs. recent European and African ancestry.
Finally, patterns of genetic similarity among inferred African seg-
ments of African-American genomes and genomes of contempo-
rary African populations included in this study suggest African
ancestry is most similar to non-Bantu Niger-Kordofanian-speaking
populations, consistent with historical documents of the African
Diaspora and trans-Atlantic slave trade.
Africa|human genomics|population genetics
identified genes under natural selection across evolutionary time
(3), and hold great potential for elucidating the genetic bases of
disease susceptibility and drug response among diverse human
populations (4, 5). The study of African population structure is
also critical for reconstructing patterns of African ancestry
among African Americans and for enabling genome-wide asso-
ciation mapping of complex disease susceptibility and pharma-
cogenomic response in African-American populations (6–9).
Africa contains over 2,000 ethnolinguistic groups and harbors
great genetic diversity (2, 10–17), but little is known about fine-
scale population structure at a genome-wide level. This is, in
part, because previous studies of high-density SNP and haplotype
variation among global human populations (defined as studies
with at least 100,000 SNP markers) have included few African
populations (10, 12, 13, 18), whereas detailed studies of genetic
structure among African populations have used a modest num-
ber of markers (2) (∼1,500 microsatellites and indels). None-
theless, recent studies of microsatellite and DNA sequence
variation suggest a significant population structure exists within
sub-Saharan Africa, with geography, language, and mode of
tudies of African genetic diversity have greatly informed our
understanding of human origins and history (1, 2), have
subsistence (e.g., hunter-gatherer, pastoralist, agriculturalist) as
potential key factors (2, 12, 13, 19). Given that high-density
genotype data have revealed discernible population structure
within other continental populations (e.g., Europe, East Asia)
and even among geographical regions within countries (e.g.,
Switzerland, Finland, United Kingdom) (20–24), there is strong
reason to believe that high-density genotype data from African
and African-American populations can elucidate patterns of
genetic structure among these populations further.
We have thus genotyped on the Affymetrix GeneChip 500K
array set 146 individuals from 11 populations in West and South
Africa (Fig. S1 and Table S1) who speak Nilo-Saharan, Afro-
Asiatic, and Niger-Kordofanian languages and integrated these
data with our previous studies of human genomic diversity, in-
cluding 57 Yorubans from Ibadan, Nigeria, genotyped as part of
the International Haplotype Map project, 365 African Ameri-
cans from throughout the United States, and 400 individuals of
European ancestry (10, 25). Our study focuses on analysis of
fine-scale population structure among the West African samples
and its implication for high-resolution inference of admixture in
African Americans. We use principal component analysis (PCA)
to infer axes of genetic variation within Africa and examine in-
dividual and population clustering using the clustering algorithm
FRAPPE (26). Next, we compare the West African, European,
and African-American samples and seek to identify the set of
West African populations closest to the ancestral population of
African Americans. Finally, based on the results of the other two
ancestry along each chromosome for each African-American
method that infers admixture proportions based on high-density
Author contributions: K.B., S.A.T., and C.D.B. designed research; K.B., A.A., S.A.T., and C.D.
B. performed research; K.B., M.R.N., J.R.O., S.L.H., S.W., A.F., J.-M.B., C.W., S.A.T., and C.D.
B. contributed new reagents/analytic tools; K.B., A.A., M.R.N., S.A.T., and C.D.B. analyzed
data; K.B., A.A., S.A.T., and C.D.B. wrote the paper; and S.A.T. and C.D.B. co-supervised
Conflict of interest statement: The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Freely available online through the PNAS open access option.
1S.A.T. and C.D.B. contributed equally to this work.
2To whom correspondence may be addressed at: Departments of Biology and Genetics,
428 Clinical Research Building, 415 Curie Boulevard, University of Pennsylvania, Philadel-
phia, PA 19104–6145. E-mail: firstname.lastname@example.org.
3To whom correspondence may be addressed at: Department of Biological Statistics,
Computational Biology, 102J Weill Hall, Cornell University, Ithaca, NY 14853. E-mail:
This article contains supporting information online at www.pnas.org/cgi/content/full/
| January 12, 2010
| vol. 107
| no. 2www.pnas.org/cgi/doi/10.1073/pnas.0909559107
Genetic Structure of West African Populations. Ourstudyfocusedon
studies suggest that region was the source for most of the ancestry
ofpresent-day African Americans (2, 27,28).Amongthe sampled
West African populations, Wright’s measure of population dif-
ferentiation [autosomal FST (29)] was low (1.2%), suggesting
quite recent common ancestry of all individuals in our sample or,
alternatively, a large effective population size for the structured
population from which the sample was drawn, with a large degree
of gene flow among subpopulations. Nonetheless, we observed
substantial variation in pairwise FSTamong sampled populations,
suggesting genetic heterogeneity among the groups (Table 1).
Differences in pairwise FST may reflect variation in effective
population size or migration rates among the populations po-
tentially attributable to isolation by distance or heterogeneity in
geographical or cultural barriers to gene flow. For example, the
Fulani appear to be genetically distinct from all other West Af-
rican populations we sampled (average pairwise FST= 3.91%).
consistently exhibited pairwise FSTabove 1% when compared
with any other population, whereas the non-Bantu Niger-
Kordofanian populations of the Igbo, Brong, and Yoruba ex-
hibited little genetic differentiation from one another (average
FST<0.4%). These results suggest that there are clear and dis-
cernible genetic differences among some of the West African
populations, whereas others appear to be nearly indistinguishable
even when comparing over 300,000 genetic markers.
To investigate whether we could reliably distinguish ancestry
among individuals from these populations, we used two ap-
proaches tailored for high-density genotype data. One, FRAPPE,
implements a maximum likelihood method to infer genetic
have originated from K ancestral clusters (26). Fig. 1A and Fig. S2
summarize FRAPPE results when the number of clusters, K, is
varied from K = 2 to K = 7. The small number of clusters was
“2” generated by tallying the number of copies of a given allele
across all SNPs in a panel for all individuals genotyped) (30).
Patterns of population structure were consistent between the
two approaches (Figs. S2–S5). For example, in the FRAPPE
analysis, the Fulani population was distinguished at K = 2, with
Bulala, Mada, and Kaba populations showing some genetic
similarity with the Fulani. PCA, likewise, separated the Fulani
from other populations along the first principal component
(PC1) (Fig. 1B). The two subsequent principal components, PC2
and PC3, reflect the geographical distribution of the populations.
PC2 showed a Chadic and Nilo-Saharan dimension extending
into inland Africa from the coast, distinguishing the Bulala,
Mada, and Kaba populations. These populations belong to the
Nilo-Saharan and Afro-Asiatic (Chadic) linguistic groups and
live further inland. Analysis of the African populations, exclud-
ing the Fulani, gave a PC1 and PC2 that resemble the second and
third principal components of the PCA with the Fulani (Fig. 1C).
Rotating the PC1 and PC2 axes from the PCA without the Fulani
reveals the similarity of the genetic and geographical maps (Fig.
1 C and D).
At K = 3, the FRAPPE algorithm clusters the Bulala into their
own group and suggests genetic similarity of the Mada, Kaba, and
Hausa,potentially indicating differentiation ofNilo-Saharan- and
Afro-Asiatic-speaking populations from Niger-Kordofanian-
speaking populations. At K = 4, all individuals from the Bantu-
speaking Xhosa of South Africa cluster into a single group and
individuals from theBantu-speaking populations (Fang,Bamoun,
and Kongo) exhibit considerable shared membership in this
cluster. At K = 5, the Mada are distinguishable as a unique group,
with modest genetic similarity with the Hausa and Kaba as well as
with most of the Niger-Kordofanian populations. These results
suggest that although these populations are quite closely related
genetically, it is possible to detect meaningful population sub-
structure given sufficient marker density [see also ref. (2)]. It is
important to note that there is likely further substructure and
diversity within these populations. Because we sample a modest
number of individuals from each population (n = 13, on average,
per population), we are not likely to have captured all the genetic
variation within each population, region, or linguistic family. To
compare patterns of haplotype structure and discern differences
in demographic history among the African populations, we esti-
mated linkage disequilibrium (LD) between all pairs of markers
in the data for all populations (see SI Text and Fig. S6). All the
African populations showed low levels of LD (even at closely
linked sites) and a rapid decay of LD with distance genome-wide
relative to populations of European ancestry.
Genome-Wide Patterns of Admixture in African Americans. To
understand the genetic structure of the African-American pop-
ulation better and to determine African-American ancestry, we
used FRAPPE to evaluate African Americans together with
European and African individuals genotyped on the same
marker set. At K = 2, African populations (blue) were dis-
tinguished from European populations (red), with African
Americans showing highly variable levels of European and West
African ancestry (Fig. 1 E and F). For the African Americans,
estimated mean West African ancestry was 77%, consistent with
prior studies (2, 28, 31–34). Analysis at K = 4 revealed additional
substructure in a North-South cline within Europe and clusters
coinciding with the linguistic and geographical substructure
within Africa (see SI Text, Tables S2–S4, and Figs. S7 and S8 for
Table 1.FSTdistances between African populations
Igbo BrongYorubaKongoBamoun XhosaFangHausaKaba MadaBulala
Bold face font is used to emphasize the genetic similarity of the Igbo, Brong and Yoruba.
Bryc et al.PNAS
| January 12, 2010
| vol. 107
| no. 2
genotype value matrix of the European, West African, and Afri-
can-American samples revealed the primary axis of variation
(see Fig. 2A) and explained ∼ 9.8% of the genetic variance. Spe-
cifically, we observed two centroids in the data, with all the in-
dividuals of European ancestry exhibiting negative loadings along
PC1, whereas all the West African individuals exhibited positive
loadings. African Americans exhibited a wide range of loadings
along PC1, presumably attributable to differences in European vs.
West African ancestry. PC2 corresponds to population sub-
structure within West Africa and largely mirrors the patterns dis-
Estimation of Admixture in Local Genomic Regions. We recon-
structed estimated European or West African ancestry for every
African American in our dataset at every position in the genome
using a PCA-based algorithm (Fig. 2A). Our method is a gen-
eralization of the approach of Paschou et al. (35) and estimates
genome-wide proportion of West African ancestry for a given
individual as P = b/(a + b), where b and a are the chord dis-
tances from the European and West African centroids, re-
spectively, for the given individual along PC1. Our generalization
involves undertaking the PC1 distance analysis on a grid of
points along the genome (as opposed to genome-wide) centered
on 15 SNP windows and using a Hidden Markov Model (HMM)
for inference of ancestry state (i.e., having “0,” “1,” or “2”
chromosomes of recent African origin; see SI Text, Fig. 2B, and
Fig. S9). An ancestry plot summarizing the number of segments
of European (i.e., “0”), West African (i.e., “2”), or admixed (i.e.,
“1”) ancestry for a representative African-American individual
a great deal of variation among the ancestry plots of the 365 self-
identified African Americans in the study, ranging from an esti-
1% West African ancestry (Fig. 2F). Some patterns reflected a
high level of West African ancestry and only one or two ancestry-
switching events per chromosome, suggesting very recent direct
African ancestry (Fig. 2D). Other patterns reflected only Euro-
pean and admixed ancestry throughout the genome, suggesting
one parent of European ancestry and one parent of African-
American ancestry (Fig. 2E).
An interesting question one can address with these kinds of
data is whether regions of the genome show substantially high
European or West African ancestry across all individuals in the
sample [e.g., as may be the case if a particular allele from one of
the ancestral populations was under strong selection (36–39)].
For our analysis, we considered genomic regions as potential
candidates for increased European or West African ancestry if
individuals was 3 SDsabove or below thegenome-wide average of
West African ancestry (78.1%). Using this approach, we found
that several genomic regions of autosomal chromosomes 5, 6, and
11 could be considered outliers from the genome-wide dis-
tribution of ancestry, although these differences were not sig-
nificant after correction for multiple tests. In Fig. 2G, we show
any outlier regions (chromosomes 1 and 12) and one chromo-
some showing a region falling outside the 3 SD criteria (chro-
mosome 11). Mean ancestry estimates for all chromosomes can
be found in Fig. S10, and a precise listing of molecular regions for
the three outlier regions may be found in Table S4. In contrast to
the autosomes, the X chromosome shows significantly high West
African ancestry along the majority of the chromosome, con-
sistent with a gender-biased model of admixture with excess Eu-
ropean male and West African female ancestry (Fig. 2G).
0.30 0.20 0.10 0.00
PC1 vs PC2
Principal Component 1 ( 1.67 %)
Principal Component 2 ( 1.02 %)
Rotated PC 1 ( 1.17 %)
0.05 0.10 0.15
PC1 vs PC2 without Fulani
Rotated PC 2 ( 0.73 %)
5 10 15 20 25
Dem. Rep. Congo
Central Af. Rep.
represented as thin vertical lines partitioned into segments corresponding to the inferred membership in K = 2 through K = 5 genetic clusters as indicated by
the colors (see Figs. S2–S5 for additional results). (B) Principal components 1 and 2 of the African individuals. (C) Principal components 1 and 2 of the African
individuals, excluding the Fulani population, wherein the components have been rotated to emphasize further similarity with geography. (D) Approximate
locations of sampled populations in Africa. (E and F) FRAPPE clustering of Europeans, African Americans, and West Africans. Individuals are represented as
thin vertical lines partitioned into K segments corresponding to the inferred membership of the genetic clusters indicated by the colors. Values for K = 2 (E)
and K = 4 (F) are shown for comparison between the two analyses.
Population structure within West Africa and relation to language and geography. (A) FRAPPE analysis of the West African populations. Individuals are
| www.pnas.org/cgi/doi/10.1073/pnas.0909559107Bryc et al.
The Bantu expansion occurred ∼4,000 years ago, originating in
Cameroon or Nigeria and expanding throughout sub-Saharan
Africa (40, 41). The clustering of the Xhosa, Fang, Bamoun, and
Kongo populations, all of which are Bantu Niger-Kordofanian-
speaking populations, likely reflects a Bantu migration from
Nigeria/Cameroon expanding toward the south. Although we
have limited sample sizes (with three of our populations having
sample sizes of less than 10), the relative order of clustering (the
East-West axis, followed by the North-South axis) suggests that
the strongest differentiating axis among the African populations
is linguistic classification corresponding to Chadic and Nilo-
Saharan vs. Niger-Kordofanian ancestry. The relatively weaker
North-South axis may result from the genetic similarity among
the Niger-Kordofanian linguistic groups because of their recent
common ancestry. Although sampled in Nigeria, the very distinct
Fulani are part of a nomadic pastoralist population that occupies
a broad geographical range across Central and Western Africa.
Analyses of microsatellite and insertion/deletion polymorphisms
indicate that they share ancestry with Niger-Kordofanian, North
African, and Central African Nilo-Saharan populations, as well
as low levels of European and/or Middle Eastern ancestry (2).
Exempting the Fulani, our LD analyses show no large differences
in rates of LD decay among our sampled African populations,
with all populations exhibiting a faster decay of LD (i.e., larger
inferred effective population size) than previously characterized
populations of European ancestry (see SI Text).
Interestingly, the Kongo population does not follow the
overall trend of East-West and North-South clustering. The
Kongo population’s genetic proximity to geographically distant
Bantu populations from Cameroon could be explained by the
genetic similarity of Bantu-speaking populations in the region, as
seen in the FRAPPE analyses (Fig. 1). Alternatively, although
these individuals self-identified as Kongo and were refugees
from locations within the Democratic Republic of Congo, the
samples were collected in Cameroon; therefore, self-identified
ancestry might poorly represent the long-term geographical
origins or may reflect recent admixture.
ordinates in PCA space (i.e., “loadings”) and the West African centroid (“a”) and the European centroid (“b”) along PC1 for PCA space that includes Eu-
ropeans, African Americans, and West Africans. (B) Local ancestry estimation using the PCA sliding window approach and associated HMM for number of
chromosomes for a given individual (i.e., “0,” “1,” or “2”) with African ancestry. (C–F) Individual ancestry estimates of 4 representative African-American
individuals (denoted 1, 2, 3, and 4 in Fig. 2A) in our dataset of 365 individuals. The colors represent two chromosomes of West African ancestry (blue), two
chromosomes of European ancestry (red), or one chromosome of West African and one chromosome of European ancestry (green). (G) Mean ancestry of 365
African-American individuals at each window across chromosome (chrom) 1, chrom 11, chrom 12, and chromosome X (X Chr). The black line shows the overall
mean estimated ancestry. Red bands indicate +3 and −3 SDs from the mean ancestry. (All chromosomes are reported in Fig. S10).
Results of our PCA-based ancestry estimation method. (A) Graphical illustration of approach: Euclidean distances from a given individual's co-
Bryc et al.PNAS
| January 12, 2010
| vol. 107
| no. 2
A concern in estimating admixture is the effect of choice of
ancestral populations. Often, the true ancestral population is no
longer available for sampling; thus, using a proxy may introduce
bias when evaluating the admixed population. For example,
individual admixture estimates in Latin Americans have been
shown to depend on the ancestral populations evaluated (42).
Some studies estimating admixture proportions in African
Americans have used a single ancestral African population, the
Yoruba (39), and our data provide an effective means of testing
whether other populations may serve as better proxies for the
ancestral population of African Americans and whether using
the Yoruba biases inferences. Comparison of the inferred West
African segments of African-American genomes with con-
temporary West African populations (Table S3) reveals that the
ancestry of the West African component of African Americans is
most similar to the profile from non-Bantu Niger-Kordofanian-
speaking populations, which include the Igbo, Brong, and Yo-
ruba, with FST values to African segments of the African
Americans ranging from 0.074 to 0.089%. That these FSTvalues
are all nearly identical (and quite small), coupled with the small
pairwise FSTvalues of the Igbo, Yoruba, and Brong populations
(Table 1), suggests that considering the set of West African
populations sampled, any of these three populations may serve
as a proxy for the ancestral population of the African Americans
and that, in fact, all three are likely to have contributed ancestry
to present-day African Americans (43). This is wholly in line with
historical documents showing that the Igbo and Yoruba are 2 of
the 10 most frequent ethnicities in slave trade records, although it
is important to note that other African populations not sampled,
including those from Sierra Leone, Senegal, Guinea Bissau, and
for the ancestral population of some African Americans (44).
That some individuals who self-identify as African American
show almost no West African ancestry and others show almost
complete West African ancestry has implications for pharma-
cogenomics studies and assessment of disease risk. Although
individuals with very low West African or very low European
ancestry may be expected by chance after several generations of
admixture, these individuals are most likely descendants of
individuals of European ancestry or recent African immigrants,
respectively. Assuming these individuals are not simply mis-
labeled, it appears that the range of genetic ancestry captured
under the term African American is extremely diverse, which
suggests caution should be used in prescribing treatment based
on differential guidelines for African Americans (45).
We found regions on chromosomes 5, 6, and 11 that show
deviations from the overall mean West African ancestry. These
regions do not overlap with those previously suggested to be
under selection (39), and about a dozen genes are found across
these regions. Whether these genes or regions are potentially
In conclusion, we believe the data presented here speak to
several important points. First, patterns of genomic diversity
are discernible using high-density genotype data and allow us to
differentiate closely related populations along linguistic and geo-
graphical axes, even with limited sample sizes from many of our
populations. Second, admixture can be reconstructed for local
genomic regions efficiently at a high density of genetic markers.
two ancestral source populations, but the approach is general-
izable to multiple populations. Application of the method to
genome-wide patterns of genomic variation in African Americans
reveals the rich mosaic structure of admixture in this population.
We find that we can distinguish African ancestry among West
Niger-Kordofanian populations) but that some populations (e.g.,
Igbo,Yoruba, and, toa lesser extent, Brong)are so closely related
genetically that their contribution to patterns of African ancestry
in African Americans is not reliably distinguishable. We believe
that increasing the density of markers and, more importantly,
sequencing directly in these populations to identify ancestry-
informative markers may make this possible in the future.
Materials and Methods
Datasets. We genotyped 225 individuals from 11 African populations [see the
article by Tishkoff et al. (2) for sampling locations] on the Affymetrix Gen-
eChip 500K array set and incorporated data from the Yoruban population of
Ibadan, Nigeria, from the HapMap project, thinned to the same SNP set (10).
European samples were from the GlaxoSmithKline Population Reference
Sample (POPRES) project, a resource of nearly 6,000 control individuals from
North America, Europe, and Asia (25) genotyped on the Affymetrix Gen-
eChip 500K array set. For our analyses, we extracted a subset of 400 in-
dividuals from Europe, randomly sampling 15 individuals from each
European country represented in POPRES when possible and 15 individuals
each from the United States, Canada, and Australia. We include 365 African
Americans from this dataset (see SI Text and ref. 25). Written informed
consent was provided by the study participants and approved by the proper
institutional review boards, and permits were obtained for collection of
African populations as described by Tishkoff et al. (2).
Population Structure Analyses. FRAPPE implements an efficient maximum
likelihood version of the Bayesian clustering algorithm, STRUCTURE (26, 46,
47). After thinning markers to have Pearson product-moment correlation of
allele frequency, r2, less than 0.5 in 50 SNP windows, shifted and recalculated
every 5 SNPs, we ran FRAPPE on all 204,457 remaining markers for 5,000
iterations. Clusters at K = 6 and higher did not correspond to known lin-
guistic or population substructures (Fig. S2). We ran PCA using the program
smartpca from the package eigenstrat (30) on a reduced dataset of 251,253
SNPs, where r2< 0.8 in 50 SNP windows. FSTwas calculated using a C++
implementation of Weir and Cockerham’s weighted equations (29). Minor
allele frequency (MAF) was thresholded at >0.1 in the populations being
compared for all comparisons, except when calculating distances between
African Americans and each of the African populations. To reduce the SNP
ascertainment biases associated with SNP discovery in the Yoruba, we used
only markers with a MAF >0.1 in Europeans for the FSTestimates.
Admixture Analysis. Our localgenomicPCAadmixturemethodnormalizesthe
genotype matrix of all individuals using the procedure as in eigenstrat (30).
Each chromosome is divided into 15 SNP nonoverlapping windows. The score
for an individual for a given window is the product of an individual’s nor-
malized and scaled genotypes across this window with the corresponding
Windows that have one or more missing genotypes for an individual are not
given a score and are omitted by the HMM. This gives a vector of scores for
each individual across all chromosomes. We assume that ancestral population
scores are drawn from a normal distribution and use the ancestral population
sample means and variances as the estimated parameters for the distribution
(see SI Text for mathematical details of the model and validation).
ACKNOWLEDGMENTS. We thank K. King for her work in managing and
preparing the POPRES data. We thank J.D. Degenhardt for helpful dis-
cussions and suggestions throughout the project, and K.E. Lohmueller for
discussion, LD scripts, and constructive comments on the manuscript. This
1R01GM83606). S.A.T. additionally acknowledges support by the National
Institutes of Health (Grant R01GM076637), National Science Foundation
(Grants BCS-0196183, BSC-0552486, and BCS-0827436), and David and Lucile
Packard and Burroughs Wellcome Foundation Career Awards.
1. Reed FA, Tishkoff SA (2006) African human diversity, origins and migrations. Curr
Opin Genet Dev 16:597–605.
2. Tishkoff SA, et al. (2009) The genetic structure and history of Africans and African
Americans. Science 324:1035–1044.
3. Tishkoff SA, et al. (2007) Convergent adaptation of human lactase persistence in
Africa and Europe. Nat Genet 39:31–40.
4. Sirugo G, et al. (2008) Genetic studies of African populations: An overview on
| www.pnas.org/cgi/doi/10.1073/pnas.0909559107 Bryc et al.
5. Campbell MC, Tishkoff SA (2008) African genetic diversity: Implications for human Download full-text
demographic history, modern human origins, and complex disease mapping. Annu
Rev Genomics Hum Genet 9:403–433.
6. Ma L, et al. (2005) Distribution of CCR2-64I and SDF1-3′A alleles and HIV status in 7
ethnic populations of Cameroon. J Acquir Immune Defic Syndr 40:89–95.
7. Williamson C, et al. (2000) Allelic frequencies of host genetic variants influencing
susceptibility to HIV-1 infection and disease in South African populations. AIDS 14:
8. Reich D, et al. (2005) A whole-genome admixture scan finds a candidate locus for
multiple sclerosis susceptibility. Nat Genet 37:1113–1118.
9. Johnson JA (2008) Ethnic differences in cardiovascular drug response: Potential
contribution of pharmacogenetics. Circulation 118:1383–1393.
10. Frazer KA, et al.; International HapMap Consortium (2007) A second generation
human haplotype map of over 3.1 million SNPs. Nature 449:851–861.
11. Garrigan D, et al. (2007) Inferring human population sizes, divergence times and rates
of gene flow from mitochondrial, X and Y chromosome resequencing data. Genetics
12. Jakobsson M, et al. (2008) Genotype, haplotype and copy-number variation in
worldwide human populations. Nature 451:998–1003.
13. Li JZ, et al. (2008) Worldwide human relationships inferred from genome-wide
patterns of variation. Science 319:1100–1104.
14. Tishkoff SA, et al. (1996) Global patterns of linkage disequilibrium at the CD4 locus
and modern human origins. Science 271:1380–1387.
15. Tishkoff SA, Kidd KK (2004) Implications of biogeography of human populations for
‘race’ and medicine. Nat Genet. 36 (11 suppl):S21–27.
16. Tishkoff SA, Verrelli BC (2003) Patterns of human genetic diversity: Implications for
human evolutionary history and disease. Annu Rev Genomics Hum Genet 4:293–340.
17. Tishkoff SA, Williams SM (2002) Genetic analysis of African populations: Human
evolution and complex disease. Nat Rev Genet 3:611–621.
18. Adeyemo AA, Chen G, Chen Y, Rotimi C (2005) Genetic structure in four West African
population groups. BMC Genet 6:38.
19. Patin E, et al. (2009) Inferring the demographic history of African farmers and pygmy
hunter-gatherers using a multilocus resequencing data set. PLoS Genet 5:e1000448.
20. Lao O, et al. (2008) Correlation between genetic and geographic structure in Europe.
Curr Biol 18:1241–1248.
21. Novembre J, et al. (2008) Genes mirror geography within Europe. Nature 456:98–101.
22. Xing J, et al. (2009) Fine-scaled human genetic structure revealed by SNP microarrays.
Genome Res 19:815–825.
23. McEvoy BP, et al. (2009) Geographical structure and differential natural selection
among North European populations. Genome Res 19:804–814.
24. Nelis M, et al. (2009) Genetic structure of Europeans: A view from the North-East.
PLoS One 4(5):e5472.
25. Nelson MR, et al. (2008) The Population Reference Sample, POPRES: A resource for
population, disease, and pharmacological genetics research. Am J Hum Genet 83:
26. Tang H, Peng J, Wang P, Risch NJ (2005) Estimation of individual admixture: Analytical
and study design considerations. Genet Epidemiol 28:289–301.
27. Lovejoy PE (2000) Transformations in Slavery (Cambridge Univ Press, New York).
28. Salas A, et al. (2005) Shipwrecks and founder effects: Divergent demographic histories
reflected in Caribbean mtDNA. Am J Phys Anthropol 128:855–860.
29. Weir BS, Cockerham CC (1984) Estimating F-statistics for the analysis of population
structure. Evolution 38:1358–1370.
30. Patterson N, Price AL, Reich D (2006) Population structure and eigenanalysis. PLoS
31. Parra EJ, et al. (2001) Ancestral proportions and admixture dynamics in geographically
defined African Americans living in South Carolina. Am J Phys Anthropol 114:18–29.
32. Lind JM, et al. (2007) Elevated male European and female African contributions to the
genomes of African American individuals. Hum Genet 120:713–722.
33. Smith MW, et al. (2004) A high-density admixture map for disease gene discovery in
African Americans. Am J Hum Genet 74:1001–1013.
34. Parra EJ, et al. (1998) Estimating African American admixture proportions by use of
population-specific alleles. Am J Hum Genet 63:1839–1851.
35. Paschou P, et al. (2007) PCA-correlated SNPs for structure identification in worldwide
human populations. PLoS Genet 3:1672–1686.
36. Workman PL, Blumberg BS, Cooper AJ (1963) Selection, gene migration and
polymorphic stability in a U.S. white and Negro population. Am J Hum Genet 15:
37. Reed TE (1969) Caucasian genes in American Negroes. Science 165:762–768.
38. Cavalli-Sforza L, Bodmer W (1971) The Genetics of Human Populations (Freeman, San
39. Tang H, et al. (2007) Recent genetic selection in the ancestral admixture of Puerto
Ricans. Am J Hum Genet 81:626–633.
40. Ehret C (2001) Bantu expansions: Re-envisioning a central problem of early African
history. Int J Afr Hist Stud 34:5–40.
41. Klieman KA (2003) The Pygmies Were Our Compass (Heinemann, Portsmouth, NH).
42. Tian C, et al. (2008) Analysis and application of European genetic substructure using
300 K SNP information. PLoS Genet. 4(1):e4. Am J Hum Genet 81(1):626–633.
43. Gabriel SB, et al. (2002) The structure of haplotype blocks in the human genome.
44. Hall GM (2005) Slavery and African Ethnicities in the Americas: Restoring the Links
(Univ North Carolina Press, Chapel Hill, NC).
45. Reiner AP, et al. (2005) Population structure, admixture, and aging-related
phenotypes in African American adults: The Cardiovascular Health Study. Am J Hum
46. Pritchard JK, Donnelly P (2001) Case-control studies of association in structured or
admixed populations. Theor Popul Biol 60:227–237.
47. Falush D, Stephens M, Pritchard JK (2003) Inference of population structure using
multilocus genotype data: Linked loci and correlated allele frequencies. Genetics 164:
Bryc et al.PNAS
| January 12, 2010
| vol. 107
| no. 2