Demographic history and rare allele sharing among human populations

Simon Gravel, Brenna M. Henn, Ryan N. Gutenkunst, Amit R. Indap, Gabor T. Marth, Andrew G. Clark, Fuli Yu, Richard A. Gibbs, The 1000 Genomes Project, Carlos D. Bustamante [email protected], David L. Altshuler, Richard M. Durbin, Gonçalo R. Abecasis, David R. Bentley, Aravinda Chakravarti, Andrew G. Clark, Francis S. Collins, Francisco M. De La Vega, Peter Donnelly, Michael Egholm, Paul Flicek, Stacey B. Gabriel, Richard A. Gibbs, Bartha M. Knoppers, Eric S. Lander, Hans Lehrach, Elaine R. Mardis, Gil A. McVean, Debbie A. Nickerson, Leena Peltonen, Alan J. Schafer, Stephen T. Sherry, Jun Wang, Richard K. Wilson, Richard A. Gibbs, David Deiros, Mike Metzker, Donna Muzny, Jeff Reid, David Wheeler, Jun Wang, Jingxiang Li, Min Jian, Guoqing Li, Ruiqiang Li, Huiqing Liang, Geng Tian, Bo Wang, Jian Wang, Wei Wang, Huanming Yang, Xiuqing Zhang, Huisong Zheng, Eric S. Lander, David L. Altshuler, Lauren Ambrogio, Toby Bloom, Kristian Cibulskis, Tim J. Fennell, Stacey B. Gabriel, David B. Jaffe, Erica Shefler, Carrie L. Sougnez, David R. Bentley, Niall Gormley, Sean Humphray, Zoya Kingsbury, Paula Koko-Gonzales, Jennifer Stone, Kevin J. McKernan, Gina L. Costa, Jeffry K. Ichikawa, Clarence C. Lee, Ralf Sudbrak, Hans Lehrach, Tatiana A. Borodina, Andreas Dahl, Alexey N. Davydov, Peter Marquardt, Florian Mertes, Wilfiried Nietfeld, Philip Rosenstiel, Stefan Schreiber, Aleksey V. Soldatov, Bernd Timmermann, Marius Tolzmann, Michael Egholm, Jason Affourtit, Dana Ashworth, Said Attiya, Melissa Bachorski, Eli Buglione, Adam Burke, Amanda Caprio, Christopher Celone, Shauna Clark, David Conners, Brian Desany, Lisa Gu, Lorri Guccione, Kalvin Kao, Andrew Kebbel, Jennifer Knowlton, Matthew Labrecque, Louise McDade, Craig Mealmaker, Melissa Minderman, Anne Nawrocki, Faheem Niazi, Kristen Pareja, Ravi Ramenani, David Riches, Wanmin Song, Cynthia Turcotte, Shally Wang, Elaine R. Mardis, Richard K. Wilson, David Dooling, Lucinda Fulton, Robert Fulton, George Weinstock, Richard M. Durbin, John Burton, David M. Carter, Carol Churcher, Alison Coffey, Anthony Cox, Aarno Palotie, Michael Quail, Tom Skelly, James Stalker, Harold P. Swerdlow, Daniel Turner, Anniek De Witte, Shane Giles, Richard A. Gibbs, David Wheeler, Matthew Bainbridge, Danny Challis, Aniko Sabo, Fuli Yu, Jin Yu, Jun Wang, Xiaodong Fang, Xiaosen Guo, Ruiqiang Li, Yingrui Li, Ruibang Luo, Shuaishuai Tai, Honglong Wu, Hancheng Zheng, Xiaole Zheng, Yan Zhou, Guoqing Li, Jian Wang, Huanming Yang, Gabor T. Marth, Erik P. Garrison, Weichun Huang, Amit Indap, Deniz Kural, Wan-Ping Lee, Wen Fung Leong, Aaron R. Quinlan, Chip Stewart, Michael P. Stromberg, Alistair N. Ward, Jiantao Wu, Charles Lee, Ryan E. Mills, Xinghua Shi, Mark J. Daly, Mark A. DePristo, David L. Altshuler, Aaron D. Ball, Eric Banks, Toby Bloom, Brian L. Browning, Kristian Cibulskis, Tim J. Fennell, Kiran V. Garimella, Sharon R. Grossman, Robert E. Handsaker, Matt Hanna, Chris Hartl, David B. Jaffe, Andrew M. Kernytsky, Joshua M. Korn, Heng Li, Jared R. Maguire, Steven A. McCarroll, Aaron McKenna, James C. Nemesh, Anthony A. Philippakis, Ryan E. Poplin, Alkes Price, Manuel A. Rivas, Pardis C. Sabeti, Stephen F. Schaffner, Erica Shefler, Ilya A. Shlyakhter, David N. Cooper, Edward V. Ball, Matthew Mort, Andrew D. Phillips, Peter D. Stenson, Jonathan Sebat, Vladimir Makarov, Kenny Ye, Seungtai C. Yoon, Carlos D. Bustamante, Andrew G. Clark, Adam Boyko, Jeremiah Degenhardt, Simon Gravel, Ryan N. Gutenkunst, Mark Kaganovich, Alon Keinan, Phil Lacroute, Xin Ma, Andy Reynolds, Laura Clarke, Paul Flicek, Fiona Cunningham, Javier Herrero, Stephen Keenen, Eugene Kulesha, Rasko Leinonen, William M. McLaren, Rajesh Radhakrishnan, Richard E. Smith, Vadim Zalunin, Xiangqun Zheng-Bradley, Jan O. Korbel, Adrian M. Stütz, Sean Humphray, Markus Bauer, R. Keira Cheetham, Tony Cox, Michael Eberle, Terena James, Scott Kahn, Lisa Murray, Aravinda Chakravarti, Kai Ye, Francisco M. De La Vega, Yutao Fu, Fiona C. L. Hyland, Jonathan M. Manning, Stephen F. McLaughlin, Heather E. Peckham, Onur Sakarya, Yongming A. Sun, Eric F. Tsung, Mark A. Batzer, Miriam K. Konkel, Jerilyn A. Walker, Ralf Sudbrak, Marcus W. Albrecht, Vyacheslav S. Amstislavskiy, Ralf Herwig, Dimitri V. Parkhomchuk, Stephen T. Sherry, Richa Agarwala, Hoda M. Khouri, Aleksandr O. Morgulis, Justin E. Paschall, Lon D. Phan, Kirill E. Rotmistrovsky, Robert D. Sanders, Martin F. Shumway, Chunlin Xiao, Gil A. McVean, Adam Auton, Zamin Iqbal, Gerton Lunter, Jonathan L. Marchini, Loukas Moutsianas, Simon Myers, Afidalina Tumian, Brian Desany, James Knight, Roger Winer, David W. Craig, Steve M. Beckstrom-Sternberg, Alexis Christoforides, Ahmet A. Kurdoglu, John V. Pearson, Shripad A. Sinari, Waibhav D. Tembe, David Haussler, Angie S. Hinrichs, Sol J. Katzman, Andrew Kern, Robert M. Kuhn, Molly Przeworski, Ryan D. Hernandez, Bryan Howie, Joanna L. Kelley, S. Cord Melton, Gonçalo R. Abecasis, Yun Li, Paul Anderson, Tom Blackwell, Wei Chen, William O. Cookson, Jun Ding, Hyun Min Kang, Mark Lathrop, Liming Liang, Miriam F. Moffatt, Paul Scheet, Carlo Sidore, Matthew Snyder, Xiaowei Zhan, Sebastian Zöllner, Philip Awadalla, Ferran Casals, Youssef Idaghdour, John Keebler, Eric A. Stone, Martine Zilversmit, Lynn Jorde, Jinchuan Xing, Evan E. Eichler, Gozde Aksay, Can Alkan, Iman Hajirasouliha, Fereydoun Hormozdiari, Jeffrey M. Kidd, S. Cenk Sahinalp, Peter H. Sudmant, Elaine R. Mardis, Ken Chen, Asif Chinwalla, Li Ding, Daniel C. Koboldt, Mike D. McLellan, David Dooling, George Weinstock, John W. Wallis, Michael C. Wendl, Qunyuan Zhang, Richard M. Durbin, Cornelis A. Albers, Qasim Ayub, Senduran Balasubramaniam, Jeffrey C. Barrett, David M. Carter, Yuan Chen, Donald F. Conrad, Petr Danecek, Emmanouil T. Dermitzakis, Min Hu, Ni Huang, Matt E. Hurles, Hanjun Jin, Luke Jostins, Thomas M. Keane, Si Quang Le, Sarah Lindsay, Quan Long, Daniel G. MacArthur, Stephen B. Montgomery, Leopold Parts, James Stalker, Chris Tyler-Smith, Klaudia Walter, Yujun Zhang, Mark B. Gerstein, Michael Snyder, Alexej Abyzov, Suganthi Balasubramanian, Robert Bjornson, Jiang Du, Fabian Grubert, Lukas Habegger, Rajini Haraksingh, Justin Jee, Ekta Khurana, Hugo Y. K. Lam, Jing Leng, Xinmeng Jasmine Mu, Alexander E. Urban, Zhengdong Zhang, Yingrui Li, Ruibang Luo, Gabor T. Marth, Erik P. Garrison, Deniz Kural, Aaron R. Quinlan, Chip Stewart, Michael P. Stromberg, Alistair N. Ward, Jiantao Wu, Charles Lee, Ryan E. Mills, Xinghua Shi, Steven A. McCarroll, Eric Banks, Mark A. DePristo, Robert E. Handsaker, Chris Hartl, Joshua M. Korn, Heng Li, James C. Nemesh, Jonathan Sebat, Vladimir Makarov, Kenny Ye, Seungtai C. Yoon, Jeremiah Degenhardt, Mark Kaganovich, Laura Clarke, Richard E. Smith, Xiangqun Zheng-Bradley, Jan O. Korbel, Sean Humphray, R. Keira Cheetham, Michael Eberle, Scott Kahn, Lisa Murray, Kai Ye, Francisco M. De La Vega, Yutao Fu, Heather E. Peckham, Yongming A. Sun, Mark A. Batzer, Miriam K. Konkel, Jerilyn A. Walker, Chunlin Xiao, Zamin Iqbal, Brian Desany, Tom Blackwell, Matthew Snyder, Jinchuan Xing, Evan E. Eichler, Gozde Aksay, Can Alkan, Iman Hajirasouliha, Fereydoun Hormozdiari, Jeffrey M. Kidd, Ken Chen, Asif Chinwalla, Li Ding, Mike D. McLellan, John W. Wallis, Matt E. Hurles, Donald F. Conrad, Klaudia Walter, Yujun Zhang, Mark B. Gerstein, Michael Snyder, Alexej Abyzov, Jiang Du, Fabian Grubert, Rajini Haraksingh, Justin Jee, Ekta Khurana, Hugo Y. K. Lam, Jing Leng, Xinmeng Jasmine Mu, Alexander E. Urban, Zhengdong Zhang, Richard A. Gibbs, Matthew Bainbridge, Danny Challis, Cristian Coafra, Huyen Dinh, Christie Kovar, Sandy Lee, Donna Muzny, Lynne Nazareth, Jeff Reid, Aniko Sabo, Fuli Yu, Jin Yu, Gabor T. Marth, Erik P. Garrison, Amit Indap, Wen Fung Leong, Aaron R. Quinlan, Chip Stewart, Alistair N. Ward, Jiantao Wu, Kristian Cibulskis, Tim J. Fennell, Stacey B. Gabriel, Kiran V. Garimella, Chris Hartl, Erica Shefler, Carrie L. Sougnez, Jane Wilkinson, Andrew G. Clark, Simon Gravel, Fabian Grubert, Laura Clarke, Paul Flicek, Richard E. Smith, Xiangqun Zheng-Bradley, Stephen T. Sherry, Hoda M. Khouri, Justin E. Paschall, Martin F. Shumway, Chunlin Xiao, Gil A. McVean, Sol J. Katzman, Gonçalo R. Abecasis, Tom Blackwell, Elaine R. Mardis, David Dooling, Lucinda Fulton, Robert Fulton, Daniel C. Koboldt, Richard M. Durbin, Senduran Balasubramaniam, Allison Coffey, Thomas M. Keane, Daniel G. MacArthur, Aarno Palotie, Carol Scott, James Stalker, Chris Tyler-Smith, Mark B. Gerstein, Suganthi Balasubramanian, Aravinda Chakravarti, Bartha M. Knoppers, Gonçalo R. Abecasis, Carlos D. Bustamante, Neda Gharani, Richard A. Gibbs, Lynn Jorde, Jane S. Kaye, Alastair Kent, Taosha Li, Amy L. McGuire, Gil A. McVean, Pilar N. Ossorio, Charles N. Rotimi, Yeyang Su, Lorraine H. Toji, Chris TylerSmith, Lisa D. Brooks, Adam L. Felsenfeld, Jean E. McEwen, Assya Abdallah, Christopher R. Juenger, Nicholas C. Clemm, Francis S. Collins, Audrey Duncanson, Eric D. Green, Mark S. Guyer, Jane L. Peterson, Alan J. Schafer, Gonçalo R. Abecasis, David L. Altshuler, Adam Auton, Lisa D. Brooks, Richard M. Durbin, Richard A. Gibbs, Matt E. Hurles, and Gil A. McVeanAuthors Info & Affiliations
Edited by Michael Lynch, Indiana University, Bloomington, IN, and approved June 3, 2011 (received for review December 24, 2010)
July 5, 2011
108 (29) 11983-11988

Abstract

High-throughput sequencing technology enables population-level surveys of human genomic variation. Here, we examine the joint allele frequency distributions across continental human populations and present an approach for combining complementary aspects of whole-genome, low-coverage data and targeted high-coverage data. We apply this approach to data generated by the pilot phase of the Thousand Genomes Project, including whole-genome 2–4× coverage data for 179 samples from HapMap European, Asian, and African panels as well as high-coverage target sequencing of the exons of 800 genes from 697 individuals in seven populations. We use the site frequency spectra obtained from these data to infer demographic parameters for an Out-of-Africa model for populations of African, European, and Asian descent and to predict, by a jackknife-based approach, the amount of genetic diversity that will be discovered as sample sizes are increased. We predict that the number of discovered nonsynonymous coding variants will reach 100,000 in each population after ∼1,000 sequenced chromosomes per population, whereas ∼2,500 chromosomes will be needed for the same number of synonymous variants. Beyond this point, the number of segregating sites in the European and Asian panel populations is expected to overcome that of the African panel because of faster recent population growth. Overall, we find that the majority of human genomic variable sites are rare and exhibit little sharing among diverged populations. Our results emphasize that replication of disease association for specific rare genetic variants across diverged populations must overcome both reduced statistical power because of rarity and higher population divergence.
The Thousand Genomes Project (1000G) is the most extensive study to date of human genomic diversity (1). The pilot phase of the project consisted of whole-genome, low-coverage sequencing of 179 samples from four HapMap populations at 2–4× coverage, an exon pilot experiment that targeted exons from over 800 genes in 697 samples across seven HapMap populations with ∼50× coverage, and a trio pilot focusing on two mother–father–child trios (1). In this article, we present an approach for combining the low-coverage and the exon pilot data, and use it to estimate the joint allele frequency spectrum for individuals of European origin in Utah (CEU), Han Chinese individuals in Beijing (CHB), Japanese individuals in Tokyo (JPT), and Yoruba individuals in Ibadan, Nigeria (YRI). Our motivation for this analysis is that two pilot projects provide complementary information: the low-coverage pilot captures most of the common variation in the populations sequenced across the accessible human genome at the cost of missing some of the rarer variants, whereas the target capture data provide a more complete picture of rare variants on an interesting subset of the data. In this article, we are interested in leveraging the strengths of the exon and low-coverage pilots to obtain accurate estimates of population genetic parameters. We will focus in particular on the P-population site frequency spectrum Φ, a P-dimensional histogram that records the joint distribution of diallelic SNPs as displayed in Fig. 1.
Fig. 1.
The two-population joint SFS from panels of Chinese individuals from Beijing (CHB) and Yoruba individuals from Ibadan, Nigeria (YRI) for variants occurring in less than 15 of 100 sequenced chromosomes in both panels. Of the 3,366 variants in the overlap of the two panels, all but 194 sites are private to a single population.
More specifically, the value Φ(f1, f2, …, fP) of bin (f1, f2, …, fP) is the number of SNPs that occurs in f1 chromosomes from population 1, f2 chromosomes from population 2, etc. Because the allele frequency in diploid population i ranges from 0 to 2ni, where ni is the number of individuals sequenced in this population, Φ is a (2n1 + 1) × (2n2 + 1) × … × (2nP + 1) array. Because the number of individuals who are successfully sequenced at any given site may vary, ni is, in practice, chosen to be somewhat smaller than the total number of individual sequenced, and each site with n > ni sequenced individuals contributes to bin f in proportion to the probability that one finds f derived alleles in a random selection of ni of the n samples.
The one-population site frequency spectrum (SFS) is a staple of population genetics and is commonly used to reveal broad patterns of selection (2, 3) and demography (35). The multiple-population SFS has received increased attention recently (610), because it provides additional information about between-population structure. Many standard population genetic statistics, such as FST and Tajima's D, are summaries of the multiple-population SFS.
In this article, we study SFSs derived from the 1000G pilot project data. We develop methods to precisely estimate SFSs from high-throughput sequencing data and use this information to estimate demographic parameters for a detailed Out-of-Africa demographic model by using ∂a∂i (6), a software package that uses diffusion approximation to calculate expected SFSs across multiple populations (6, 10). We use these parameters to predict the number of variants to be discovered as the number of sequenced samples in the 1000G is increased. We also present a jackknife-based approach to the prediction of the number of undiscovered variants and compare the predictions of the two approaches.

Theory

Linear Error Model for SFSs from Low-Coverage Data.

Data of the kind collected by the 1000G low-coverage pilot (1) (2–4× coverage across 179 individuals) provide a large volume of data from which precise demographic inference can be drawn. However, the low coverage leads to biases that must be addressed to ensure accuracy of the inference (11, 12). We use an empirical approach to tune an error model for low-coverage sequencing based on a direct comparison of the SFSs generated by whole genome and capture experiments on the part of the genome sequenced by both experiments.
The usefulness of this approach relies on two observations. First, demographic inference does not require the knowledge of the particular sites that are variable, but rather requires statistical averages over all sites. Although it is impossible to infer which variable sites were missed, the average number of missed sites can be estimated directly. Second, the most significant bias caused by low coverage in the 1000G data is an elevated false-negative rate for rare variant genotype calls (1). Because the majority of genetic variants are rare, it is possible to infer error rates for such variants based on high-quality sequence data from a relatively small subset of the genome.
Because the SFS does not keep track of linkage information, we use an error model that acts independently on each genomic site. We suppose that the underlying true SFS and the observed SFS Si for population p are related by a linear error model: . In this model, represents the proportion of sites with true frequency i′ that are assigned to frequency i. In the three-population case, which we will consider below, this model generalizes to
If we have Nc frequency bins per population, the number of parameters in this model is . We therefore need to introduce additional simplifying assumptions (justifications are provided below):
i) The errors occur independently in each population: .
ii) The probability of missing a site decays exponentially with the number of variants present in the population:
iii) If a site is found to be variable in one population, its frequency is estimated accurately:
The resulting model has six parameters (amplitudes αp and error decay rates βp) and captures the bulk of the discrepancy between the high- and low-coverage data. It is motivated by the following observations:
i) The low-coverage SNP calls were made population by population. We assume, first, that the leading source of error is an insufficient number of variant reads to confidently call a variant and, second, that in a low-coverage experiment, the uncorrelated sampling fluctuations in read numbers play the largest role in the variation in read numbers.
ii) Variant calls in 1000G require multiple independent observations of a variant across a population to rule out read errors and call a variant genotype. This stringency strongly reduces the rate of false-positive calls, but it results in missing actual variants at a rate that depends on the expected number of nonreference reads observed at a given position across a population (1). The decay in the probability of detecting less than a fixed, small number c of reads for a variant present in i of N chromosome sequenced at depth d is dominated by . We therefore fit a heuristic exponential error model αe−βi, where the effective read depth 2β accounts for fluctuations in read depth and read quality across the genome (Table 1).
iii) After enough variant reads are found in a 1000G population to justify a variant call, a single variant read is sufficient to make a genotype variant call, hence a reduced false-negative rate and low systematic bias in estimated frequencies when a variant has been identified.
Table 1.
Parameter values in the error model (Eq. 1)
pαpβpMean mapped depth
CEU1.4660.7374.62×
CHB + JPT1.8550.7542.65×
YRI1.2200.5513.42×
Note that the parameter values differ substantially between populations and that error rates decrease with increased coverage.

Estimating the Parameters of the Error Model.

Because errors are assumed to occur independently in each population, the error rates can be inferred directly from error rates in single-population SFSs. We compute the single-population SFSs at sites that are found to be variable in the exon pilot data and with at least 80 genotype calls in all three populations. The exon pilot and low-coverage pilot are then compared, and the optimal parameters αp and βp are obtained through a linear fit using the first three frequency bins of the compared spectra.
Note that this error model can be inverted to give a correction model for the SFS, which does not require the knowledge of the number of fixed variants (SI Appendix). However, the correction model may involve the subtraction of large numbers and has non-Poisson uncertainties. When inferring demographic parameters by maximum likelihood of a Poisson Random Field (6), we therefore incorporate the error model in our demographic model rather than attempt to correct the SFS.

Prediction of the Rate of Variant Discovery.

One practical use of inferring a demographic model is the ability to predict the number of variants that will be discovered in subsequent experiments. To study the impact of model choice on such predictions, we propose an alternate predictor of discovery rate based on sampling theory and inspired by an analogy with capture–recapture approaches to estimating animal population sizes (1315) (let us consider rabbits for definiteness). In this analogy, a rabbit is akin to a SNP, a field trip is akin to an individual sequenced, and a rabbit capture is akin to the identification of a variant in a sequenced individual. In the absence of measurement errors, the probability of identifying a variant in a randomly chosen sequenced individual is proportional to the frequency of the variant in the population. This distribution of probabilities is akin to the variability in rabbit capture probability; a common SNP is akin to a trap-happy rabbit, and a rare SNP is akin to a trap-shy rabbit.
We propose a population genetics analog of the Burnham–Overton jackknife (16, 17) to estimate the total number V(N) of segregating sites in a sample of N chromosomes based on a subsample of n-sequenced chromosomes. This jackknife estimator uses the assumption that
where for a fixed jackknife order p. Explicit expressions for the as well as performance benchmarking and additional discussion of this estimator are provided in SI Appendix.

Results

After filtering away the exon pilot calls based on less than 15× coverage and individuals with substantial discrepancy with HapMap (SI Appendix), we compared the joint SFSs with the expected spectra obtained if each individual had been assigned to a population randomly in the independent sites model. That is, given an N × N spectrum φ(i, j), we have an expected spectrum of
Figs. 2 and 3 indicate that, even for pairs of closely related populations, we find a substantial reduction in allele sharing for rare variants compared with a single randomly mixing population. In particular, Fig. 2, Right shows the Anscombe residuals between expectation and data. Blue strips along the axis correspond to a significant excess of variants private to one panel in the data, and they are accompanied by a reduction in shared variants (red). The residuals are larger (darker colors) for rare variants not only because of a larger number of sites but also because of reduced sharing. Indeed, we see in Fig. 3 that the amount of sharing, expressed as a proportion of the expectation in a panmictic population, is only a few percent between continental populations for variants present at 2% minor allele frequency (MAF) and about 60% for variants at 20% MAF. More closely related populations, such as CHB and JPT, still exhibit a 50% reduction in sharing at 2% MAF but barely any reduction for variants at 20% MAF. Interestingly, even closely related populations, such as CHB and CHD, exhibit a 20% reduction in sharing for 2% MAF. This finding is consistent with recent population structure, but although this analysis used only genotype calls with high-coverage data, such a reduction in sharing could also be partly explained by differences in the sequencing platform between the two populations.
Fig. 2.
Joint allele SFSs (all sites) for selected pairs of populations from the exome sequencing panel (Left) compared with expected spectra under site by site population label permutation (Center). Shown are sites occurring in, at most, 15 of 100 chromosomes. All population pairs, including two different panels of Chinese individuals sampled in Beijing, China and Denver, Colorado (CHB and CHD), as well as two groups of European origin (CEU and Tuscans from Italy, or TSI), show substantial residuals for rare variants (Right), consistent with reduced sharing. White bins contain less than one count.
Fig. 3.
The probability that two individuals carrying an allele of given minor frequency come from different populations, normalized by the expected frequency in a panmictic population, using the seven panels of the exome capture dataset. Sharing decreases dramatically as frequency approaches zero. The reduction in sharing at 50% frequency in some population pairs is caused by low overall numbers in that bin, and a single site (rs6662929) that exhibits inconsistent calls between different calling platforms and most likely has an incorrect homozygous reference call in some populations. Sites were binned by frequency: dots indicate the center of each bin, and solid lines are to guide the eye. Note that singletons are not shown, because there can be no sharing for such sites.
To increase the number of sites available for estimating joint SFS, we turned to the low-coverage pilot data. Direct comparison of low-coverage and exon capture genotype calls at sites called in the exon capture pilot shows a significant discrepancy for rare variants (SI Appendix, Figs. S1–S3) because of elevated rates of false-negative variant calls in low-coverage data. This finding results in biased estimates of the distribution of allele frequencies.
The bulk of the systematic discrepancy between high- and low-coverage SNP calls could be described using the simple false-negative model described above (Table 1 and SI Appendix, Figs. S1–S4). The most substantial discrepancy between this model and the data is in the CHB + JPT, possibly because this group was the metapopulation with the lowest coverage. In this case, the high-coverage singleton counts are 634% higher than the uncorrected low-coverage counts. After error correction, a discrepancy of 19% remains, with the corrected low-coverage site predicting more counts (SI Appendix, Fig. S3). Despite the high false-negative rate for singleton calls in the low-coverage data, its sheer volume provides an advantage in estimation precision over the much smaller exon pilot dataset. Similarly, the false-negative model for multiple-population SFSs (Eq. 1) was found to account for the bulk of the discrepancy between the multiple-population SFS derived from low- and high-coverage SNP calls (SI Appendix, Fig. S4).
We modeled the joint SFS for synonymous sites in African (YRI), Asian (CHB and JPT), and European (CEU) data sequenced in the low-coverage pilot using the 13-parameter demographic model used in ref. 6 (Fig. 4 and Table 2), taking into account the expected error model. The SFS was calculated using n = 40 samples per panel (80 chromosomes). We obtained maximum composite likelihood estimates for the 13 parameters using ∂a∂i, a diffusion-approximation-based package for estimating expected SFSs resulting from various demographic models (6) (Table 2). Our maximum likelihood parameters are broadly consistent with previously reported values using National Institute on Environmental Health Sciences (NIEHS) data (6). However, the resulting confidence intervals, determined by conventional bootstrap (likelihood profiles are provided in SI Appendix), are substantially narrower than those intervals resulting from NIEHS or high-coverage data alone. As an example, using a 25 y generation time, we find a time of split between African and Eurasian populations of TB = 51 thousand years ago (kya; 95% confidence interval = 45–69 kya). By contrast, the NIEHS data (6) resulted in a maximum likelihood estimate of TB = 140 kya (95% confidence interval = 40–270 kya). The inference based on the exon pilot alone yields TB = 98 kya (95% confidence interval = 43–210 kya). In general, the gain in precision was strongest for the parameters involved in more ancient events. Inference based on uncorrected low-coverage data yielded an unrealistic TB = 14 kya split.
Fig. 4.
An illustration of the inferred demographic model, with line width corresponding to population size and time flowing from left to right. The width of the red arrows is proportional to the migration intensity. Model details are provided in Table 2 and ref. 6.
Table 2.
Parameter estimates obtained using the NIEHS data (6), 1000G exon and low-coverage data, and 1000G exon pilot data only (this work)
 NIEHSLow-coverage + exonsExons
ParameterEstimate95% CIEstimate95% CIEstimate95% CI
NA7,3004,400–10,1007,3106,984–7,7397,3103,647–9,208
NAF12,30011,500–13,90014,47413,419–16,18415,38814,240–17,542
NB2,1001,400–2,9001,8611,453–2,4942,758896–3,450
NEUO1,000500–1,9001,032677–1,2901,620991–2,759
rEU (%)0.400.15–0.660.380.28–0.590.270.17–0.39
NAS0510310–910554376–813821616–1,226
rAS (%)0.550.23–0.880.480.30–0.750.310.18–0.42
mAF B (×10−5)2515–341512–19205.5–29.6
mAF EU (×10−5)3.02.0–6.02.52.1–3.11.71.0–2.8
mAF AS (×10−5)1.90.3–10.40.780.4–1.20.580.23–1.24
mEU AS (×10−5)9.62.3–17.43.111.8–3.95.94.1–8.2
TAF (kya)220100–510148114–183316155–545
TB (kya)14040–2705145–699843–210
TEU − AS (kya)21.217.2–26.52321–272823–38
The 95% confidence intervals (CIs) in this work are obtained by bootstrap over coding regions. The equality of the maximum likelihood values for NA in the corrected low-coverage + exons data is a consequence of a normalization of the effective sequenced length (details in Methods). The resulting demographic model is illustrated in Fig. 4.
Beyond their fundamental interest as descriptors of human history, these parameters allow for a number of experimental predictions; given a demographic model, we can predict, for example, the number of synonymous variants to be discovered in samples of larger size that are currently in the process of being sequenced. We predicted the number of variants to be discovered in each of the three population considered (CEU, CHB + JPT, and YRI) as the sample size is increased using both the inferred demographic model and the jackknife estimator of the number of undiscovered variants presented in Methods (Fig. 5). Because the jackknife does not rely on assumptions about demography and selection, we also used it to predict the number of nonsynonymous sites to be discovered. The jackknife approach predicts that, as sample size is increased, the total number of segregating sites in CEU and CHB + JPT panels should overtake the number of segregating sites in the YRI population.
Fig. 5.
Observed and projected numbers of synonymous and nonsynonymous variants in CEU, CHB + JPT, and YRI as a function of the sample size (two times the number of individuals sequenced). Long and short dashes correspond to jackknife and model-based projections for synonymous sites, respectively. The dotted lines are jackknife projections for nonsynonymous sites. Discrepancies between projections are accounted for by the difference between the model prediction and observed number of singletons in the data.

Discussion

Our results illustrate that the vast majority of human variable sites are rare and that the majority of rare variants exhibit, at most, very little sharing among continental populations. We also find reduced sharing for rare variants compared with common variants among more closely related populations, such as CHB-JPT, CEU-TSI, and CHB-CHD. This lack of sharing can be explained by population divergence, and we expect that the fraction of newly discovered variable sites that are population-specific will keep increasing with sample size. This finding poses a formidable challenge for the reproduction of genome-wide association studies for rare functional variants across diverse populations, because the statistical difficulties caused by variant rarity within a population combine with increased between-population divergence.
We also show how sequencing a large number of individuals at low coverage is an efficient strategy not only for discovering the maximum number of variable sites but also for estimating demographic parameters, at least when error rates can be estimated. Different statistical methods have been proposed that include read depth information and models of sequencing errors to reduce biases in allele frequency estimation (11, 12). Because of the availability of high-coverage data for a subset of the genome in the populations studied here, we used direct comparison with high-coverage data to estimate and correct biases caused by low coverage. A significant advantage of the direct comparison approach is simplicity and computational efficiency; it can use existing curated genotype calls rather than require a full analysis of an error model at the individual read level. This advantage is particularly useful for data generated by 1000G, because multiple sequencing platforms and calling pipelines with different error modes have been used jointly. In general, the two approaches are not mutually exclusive, and when practical, a statistically corrected low-coverage SFS could be further corrected by comparison with targeted high-coverage data. Here, we used, as a reference, an exon capture dataset with >50× coverage and validation rates of 96.8% overall and 93.8% for singletons. We also restricted our analysis to a high-quality subset of the data (by selecting individuals with good coverage and HapMap concordance and selecting sites with sufficient coverage). The false-negative rate in the exon capture data was estimated to be below 5% for variants of at least 1% in frequency and 26% for variants below 1% in frequency. To avoid resulting biases in the frequency-dependent false-negative estimates for the low-coverage data, we restricted the comparison with sites where a high-coverage variant call had been made.
We found that the bulk of the discrepancy between high- and low-coverage data could be described by a simple model that uses only two parameters per population, and in which error rates decay exponentially with MAF, frequencies of detected variant sites are accurately determined, and errors occur independently in each population. The latter assumption is perhaps the most debatable: we expect at least some correlations in the coverage at a given site for different populations. An error model taking into account such correlations would, therefore, be desirable. However, given the limited data available to infer the parameters of the error model, the independence assumption is a reasonable tradeoff that allows for the capture of the bulk of the error patterns. Finally, the error rates likely differ between different genomic regions (such as coding vs. noncoding DNA), motivating our focus on exonic regions where high-coverage data were available. This finding emphasizes the importance of obtaining high-quality genotype data through sequencing or chip genotyping for representative noncoding regions.
The demographic model discussed in this paper was introduced in Gutenkunst et al. (6), where it was used to analyze the NIEHS intergenic data. Despite differences in putatively neutral sites (selected intergenics vs. synonymous), sequencing technology (Sanger vs. high throughput), and panel choice (CHB only vs. CHB + JPT), the inferred parameters are in broad agreement (Table 2). Inference based only on capture data provides overlapping 95% confidence intervals, with the single exception of Europe–Asia migration rate (1.8 − 3.9 × 10−5 vs. 4.1 − 8.2 × 10−5). The main difference between these three sets of parameter estimates is the width of the confidence intervals. The inference based on exon capture data provides reduced uncertainty compared with the NIEHS data, despite a comparable number of variable sites in the SFS; the additional number of samples per site results in more accurate frequency estimates that further constrain the demographic model. A much greater reduction in the confidence intervals is obtained by considering the low-coverage and exon capture data jointly (a 90% reduction of the confidence interval for the Out-of-Africa split time compared with a 27% reduction with the exon data only). Our estimate of the Out-of-Africa split time using the low-coverage data, 51 kya, is also in better agreement with both prior genetic and archaeological estimates of the modern human expansion out of Africa (18). It should be emphasized that, because we use a single Western African population as our African panel, the divergence described by our model might have occurred earlier than the actual Out-of-Africa event.
The narrow confidence intervals on some of the parameters should not obscure the fact that the parameter estimates are model-dependent. As a simple example, a model that does not allow for migration would require more recent split times to produce similar levels of population divergence. The demographic history of the four populations considered is much more eventful than what is accounted for by our model. Additional geographically intermediate populations from the Near East and Central Asia that were not included in our analysis might contribute significantly to the allele frequency distribution as ghost populations (19). Incorporating an appropriate number of source populations for estimates of migration has been a general limitation of two- and three-population models under isolation migration coalescent, approximate Bayesian computation, and diffusion-based approaches. This limitation might explain why our estimate of the divergence between East Asians and Europeans is more recent than estimates based on archaeological evidence (18), but is comparable with estimates of 23 kya (20) under an approximate Bayesian computation approach and 25 kya under an isolation migration approach with mtDNA X and Y sequence data (21).
Similarly, the current population sizes inferred from our model (15,500, 35,900, and 49,000 for YRI, CEU, and CHB, respectively) are still significantly lower than census sizes. Because our model accounts for some population size changes, these are expected to be in closer relationship to census sizes compared with the classical effective population size, but additional model refinement [such as structure within populations, generation overlap, and a recent increase in growth rate, which was observed in the work by Coventry et al. (22), in a sample of 10,422 European-Americans] will be needed to close the gap.
Predictions based on the demographic model and the jackknife approach differ as to the number of new variants to be discovered, particularly for CHB + JPT (Fig. 5). This difference is easily understood by considering the differences in the two approaches. The demographic model attempts to fit the complete SFS at the cost of model assumptions that might bias the results. By contrast, the jackknife approach focuses on the rare variants, and the model assumptions are weaker. The difference can be traced to the fact that the maximum likelihood demographic model predicts a number of singletons somewhat lower than the observed number (SI Appendix, Fig. S5). If this discrepancy is due to limitations in the model that fail to account for an excess of rare variants, we expect the jackknife estimator to be more accurate. By contrast, if the difference is because of inaccurate singleton frequency estimation (from sequencing errors leading to 6.2% of false-positive variants in the high-coverage data) or limitations of our correction model (SI Appendix, Figs. S1–S3), the demographic model is expected to provide more robust estimates.
Nonetheless, both methods predict at least 50,000 synonymous variants in the human genome when sequencing 1,000 individuals for the CEU and CHB populations, substantially more than would be predicted from population genetic models of constant size. The jackknife approach applied directly to the seven target capture populations shows similar patterns, with some variation within continents, in JPT samples showing less rare variants than the Chinese populations, and in TSI samples showing more rare variants than CEU (SI Appendix, Fig. S7). These results highlight the importance, for the planning of medical sequencing experiments, of accurate demographic models of human populations and the dramatic impact that recent human population growth has had on the structure of genetic variation. Specifically, our prediction that most genetic variants are rare and highly diverged suggests that genome-wide association studies aiming to correlate common disease susceptibility with rare variants may need extraordinarily large sample sizes and precise definitions of population samples to accurately compare frequencies in cases and controls. Eventually, a clear tradeoff will ensue between cataloging variants and genotyping vs. completely sequencing human genomes and comparing them among populations of cases and controls.

Methods

Numerics.

The unprecedented size of the 1,000 genomes data created challenges for the numerical solution of the diffusion equation. Namely, the number of grid points required to accurately estimate the three population SFS grows rapidly with the number of samples in each population. We optimized ∂a∂i and released version 1.5.0, in which the number of grid points necessary to achieve a given accuracy is reduced. As in ref. 6, we obtained SFSs with three different grid sizes (60, 70, and 80) and extrapolated to infinite grid size. Each likelihood evaluation took between 1 and 2 min on a 2.26-GHz processor. Optimization required hundreds to thousands of likelihood evaluations. Likelihoods were computed using the folded SFS to avoid biases caused by ancestral misidentification. Convergence of the maximum likelihood optimization process was ensured by restarting the search with modified initial conditions. The maximal likelihood parameters were chosen, but differences in parameter estimates from the different restarts were, on average, much smaller than the reported confidence intervals.

Conversion from Genetic to Physical Units.

The different parameters involved in the diffusion equation solved by ∂a∂i are normalized by the ancestral population size Na during the likelihood maximization. The optimal value of Na is calculated using the fact that the total number of segregating sites in a sample of n individuals is proportional to NaLμ, where μ is the mutation rate and L is the effective length sequenced. For this analysis, we used L = 5,007,837, the number of autosomal fourfold degenerate sites that passed quality control in all three populations, and μ = 2.36 × 10−8. For estimates based on exon data alone, we fixed the effective sequencing length to 68% of the target length by requesting equal values for NA in the corrected low-coverage and exon pilot estimates. The remaining 32% is composed of called sites that failed quality controls and sites for which no genotype call has been made. When performing bootstrap analysis, the total number of fourfold degenerate sites varied from bootstrap sample to bootstrap sample and was adjusted accordingly. Finally, to convert generation time to years, we used a generation time of 25 y. Estimated parameters are shown in Table 2.

Supporting Information

Appendix (PDF)
Supporting Information
sapp01.docx

References

1
; 1000 Genomes Project Consortium, A map of human genome variation from population-scale sequencing. Nature 467, 1061–1073 (2010).
2
AR Boyko, et al., Assessing the evolutionary impact of amino acid mutations in the human genome. PLoS Genet 4, e1000083 (2008).
3
SH Williamson, et al., Simultaneous inference of selection and population growth from patterns of variation in the human genome. Proc Natl Acad Sci USA 102, 7882–7887 (2005).
4
AM Adams, RR Hudson, Maximum-likelihood estimation of demographic parameters using the frequency spectrum of unlinked single-nucleotide polymorphisms. Genetics 168, 1699–1712 (2004).
5
GT Marth, E Czabarka, J Murvai, ST Sherry, The allele frequency spectrum in genome-wide human variation data reveals signals of differential demographic history in three large world populations. Genetics 166, 351–372 (2004).
6
RN Gutenkunst, RD Hernandez, SH Williamson, CD Bustamante, Inferring the joint demographic history of multiple populations from multidimensional SNP frequency data. PLoS Genet 5, e1000695 (2009).
7
R Nielsen, et al., Darwinian and demographic forces affecting human protein coding genes. Genome Res 19, 838–849 (2009).
8
SA Sawyer, DL Hartl, Population genetics of polymorphism and divergence. Genetics 132, 1161–1176 (1992).
9
CD Bustamante, J Wakeley, S Sawyer, DL Hartl, Directional selection and the site-frequency spectrum. Genetics 159, 1779–1788 (2001).
10
X Yi, et al., Sequencing of 50 human exomes reveals adaptation to high altitude. Science 329, 75–78 (2010).
11
M Lynch, Estimation of allele frequencies from high-coverage genome-sequencing projects. Genetics 182, 295–301 (2009).
12
Y Li, et al., Resequencing of 200 human exomes identifies an excess of low-frequency non-synonymous coding variants. Nat Genet 42, 969–972 (2010).
13
I Ionita-Laza, C Lange, N M Laird, Estimating the number of unseen variants in the human genome. Proc Natl Acad Sci USA 106, 5008–5013 (2009).
14
I Ionita-Laza, NM Laird, On the optimal design of genetic variant discovery studies. Stat Appl Genet Mol Biol 9, 33 (2010).
15
J Bunge, M Fitzpatrick, Estimating the number of species. A review. J Am Stat Assoc 88, 364–373 (1993).
16
K Burnham, W Overton, Estimation of the size of a closed population when capture probabilities vary among animals. Biometrika 65, 625–633 (1978).
17
K Burnham, W Overton, Robust estimation of population size when capture probabilities vary among animals. Ecology 60, 927–936 (1979).
18
RG Klein, JJ Hublin The Human Career. Human Biological and Cultural Origins (University of Chicago Press, Chicago, 1999).
19
P Beerli, Effect of unsampled populations on the estimation of population sizes and migration rates between sampled populations. Mol Ecol 13, 827–836 (2004).
20
G Laval, E Patin, LB Barreiro, L Quintana-Murci, Formulating a historical and demographic model of recent human evolution based on resequencing data from noncoding regions. PLoS One 5, e10284 (2010).
21
D Garrigan, et al., Inferring human population sizes, divergence times and rates of gene flow from mitochondrial, X and Y chromosome resequencing data. Genetics 177, 2195–2207 (2007).
22
A Coventry, et al., Deep resequencing reveals excess rare recent variants consistent with explosive population growth. Nat Commun 1, 131 (2010).

Information & Authors

Information

Published in

Go to Proceedings of the National Academy of Sciences
Proceedings of the National Academy of Sciences
Vol. 108 | No. 29
July 19, 2011
PubMed: 21730125

Classifications

Submission history

Published online: July 5, 2011
Published in issue: July 19, 2011

Keywords

  1. demographic inference
  2. genetic drift
  3. population genetics
  4. human evolution

Notes

This article is a PNAS Direct Submission.

Authors

Affiliations

Simon Gravel
Department of Genetics, Stanford University School of Medicine, Stanford, CA 94305-5120;
Brenna M. Henn
Department of Genetics, Stanford University School of Medicine, Stanford, CA 94305-5120;
Ryan N. Gutenkunst
Department of Molecular and Cellular Biology, University of Arizona, Tucson, AZ 85721;
Amit R. Indap
Department of Biology, Boston College, Chestnut Hill, MA 02467;
Gabor T. Marth
Department of Biology, Boston College, Chestnut Hill, MA 02467;
Andrew G. Clark
Department of Molecular Biology and Genetics, Cornell University, Ithaca, NY 14853; and
Fuli Yu
Human Genome Sequencing Center, Baylor College of Medicine, Houston, TX 77030
Richard A. Gibbs
Human Genome Sequencing Center, Baylor College of Medicine, Houston, TX 77030
The 1000 Genomes Project
Human Genome Sequencing Center, Baylor College of Medicine, Houston, TX 77030
Carlos D. Bustamante1 [email protected]
Department of Genetics, Stanford University School of Medicine, Stanford, CA 94305-5120;
David L. Altshuler
The 1000 Genomes Project
Richard M. Durbin
The 1000 Genomes Project
Gonçalo R. Abecasis
The 1000 Genomes Project
David R. Bentley
The 1000 Genomes Project
Aravinda Chakravarti
The 1000 Genomes Project
Andrew G. Clark
The 1000 Genomes Project
Francis S. Collins
The 1000 Genomes Project
Francisco M. De La Vega
The 1000 Genomes Project
Peter Donnelly
The 1000 Genomes Project
Michael Egholm
The 1000 Genomes Project
Paul Flicek
The 1000 Genomes Project
Stacey B. Gabriel
The 1000 Genomes Project
Richard A. Gibbs
The 1000 Genomes Project
Bartha M. Knoppers
The 1000 Genomes Project
Eric S. Lander
The 1000 Genomes Project
Hans Lehrach
The 1000 Genomes Project
Elaine R. Mardis
The 1000 Genomes Project
Gil A. McVean
The 1000 Genomes Project
Debbie A. Nickerson
The 1000 Genomes Project
Leena Peltonen
The 1000 Genomes Project
Alan J. Schafer
The 1000 Genomes Project
Stephen T. Sherry
The 1000 Genomes Project
Jun Wang
The 1000 Genomes Project
Richard K. Wilson
The 1000 Genomes Project
Richard A. Gibbs
The 1000 Genomes Project
David Deiros
The 1000 Genomes Project
Mike Metzker
The 1000 Genomes Project
Donna Muzny
The 1000 Genomes Project
Jeff Reid
The 1000 Genomes Project
David Wheeler
The 1000 Genomes Project
Jun Wang
The 1000 Genomes Project
Jingxiang Li
The 1000 Genomes Project
Min Jian
The 1000 Genomes Project
Guoqing Li
The 1000 Genomes Project
Ruiqiang Li
The 1000 Genomes Project
Huiqing Liang
The 1000 Genomes Project
Geng Tian
The 1000 Genomes Project
Bo Wang
The 1000 Genomes Project
Jian Wang
The 1000 Genomes Project
Wei Wang
The 1000 Genomes Project
Huanming Yang
The 1000 Genomes Project
Xiuqing Zhang
The 1000 Genomes Project
Huisong Zheng
The 1000 Genomes Project
Eric S. Lander
The 1000 Genomes Project
David L. Altshuler
The 1000 Genomes Project
Lauren Ambrogio
The 1000 Genomes Project
Toby Bloom
The 1000 Genomes Project
Kristian Cibulskis
The 1000 Genomes Project
Tim J. Fennell
The 1000 Genomes Project
Stacey B. Gabriel
The 1000 Genomes Project
David B. Jaffe
The 1000 Genomes Project
Erica Shefler
The 1000 Genomes Project
Carrie L. Sougnez
The 1000 Genomes Project
David R. Bentley
The 1000 Genomes Project
Niall Gormley
The 1000 Genomes Project
Sean Humphray
The 1000 Genomes Project
Zoya Kingsbury
The 1000 Genomes Project
Paula Koko-Gonzales
The 1000 Genomes Project
Jennifer Stone
The 1000 Genomes Project
Kevin J. McKernan
The 1000 Genomes Project
Gina L. Costa
The 1000 Genomes Project
Jeffry K. Ichikawa
The 1000 Genomes Project
Clarence C. Lee
The 1000 Genomes Project
Ralf Sudbrak
The 1000 Genomes Project
Hans Lehrach
The 1000 Genomes Project
Tatiana A. Borodina
The 1000 Genomes Project
Andreas Dahl
The 1000 Genomes Project
Alexey N. Davydov
The 1000 Genomes Project
Peter Marquardt
The 1000 Genomes Project
Florian Mertes
The 1000 Genomes Project
Wilfiried Nietfeld
The 1000 Genomes Project
Philip Rosenstiel
The 1000 Genomes Project
Stefan Schreiber
The 1000 Genomes Project
Aleksey V. Soldatov
The 1000 Genomes Project
Bernd Timmermann
The 1000 Genomes Project
Marius Tolzmann
The 1000 Genomes Project
Michael Egholm
The 1000 Genomes Project
Jason Affourtit
The 1000 Genomes Project
Dana Ashworth
The 1000 Genomes Project
Said Attiya
The 1000 Genomes Project
Melissa Bachorski
The 1000 Genomes Project
Eli Buglione
The 1000 Genomes Project
Adam Burke
The 1000 Genomes Project
Amanda Caprio
The 1000 Genomes Project
Christopher Celone
The 1000 Genomes Project
Shauna Clark
The 1000 Genomes Project
David Conners
The 1000 Genomes Project
Brian Desany
The 1000 Genomes Project
Lisa Gu
The 1000 Genomes Project
Lorri Guccione
The 1000 Genomes Project
Kalvin Kao
The 1000 Genomes Project
Andrew Kebbel
The 1000 Genomes Project
Jennifer Knowlton
The 1000 Genomes Project
Matthew Labrecque
The 1000 Genomes Project
Louise McDade
The 1000 Genomes Project
Craig Mealmaker
The 1000 Genomes Project
Melissa Minderman
The 1000 Genomes Project
Anne Nawrocki
The 1000 Genomes Project
Faheem Niazi
The 1000 Genomes Project
Kristen Pareja
The 1000 Genomes Project
Ravi Ramenani
The 1000 Genomes Project
David Riches
The 1000 Genomes Project
Wanmin Song
The 1000 Genomes Project
Cynthia Turcotte
The 1000 Genomes Project
Shally Wang
The 1000 Genomes Project
Elaine R. Mardis
The 1000 Genomes Project
Richard K. Wilson
The 1000 Genomes Project
David Dooling
The 1000 Genomes Project
Lucinda Fulton
The 1000 Genomes Project
Robert Fulton
The 1000 Genomes Project
George Weinstock
The 1000 Genomes Project
Richard M. Durbin
The 1000 Genomes Project
John Burton
The 1000 Genomes Project
David M. Carter
The 1000 Genomes Project
Carol Churcher
The 1000 Genomes Project
Alison Coffey
The 1000 Genomes Project
Anthony Cox
The 1000 Genomes Project
Aarno Palotie
The 1000 Genomes Project
Michael Quail
The 1000 Genomes Project
Tom Skelly
The 1000 Genomes Project
James Stalker
The 1000 Genomes Project
Harold P. Swerdlow
The 1000 Genomes Project
Daniel Turner
The 1000 Genomes Project
Anniek De Witte
The 1000 Genomes Project
Shane Giles
The 1000 Genomes Project
Richard A. Gibbs
The 1000 Genomes Project
David Wheeler
The 1000 Genomes Project
Matthew Bainbridge
The 1000 Genomes Project
Danny Challis
The 1000 Genomes Project
Aniko Sabo
The 1000 Genomes Project
Fuli Yu
The 1000 Genomes Project
Jin Yu
The 1000 Genomes Project
Jun Wang
The 1000 Genomes Project
Xiaodong Fang
The 1000 Genomes Project
Xiaosen Guo
The 1000 Genomes Project
Ruiqiang Li
The 1000 Genomes Project
Yingrui Li
The 1000 Genomes Project
Ruibang Luo
The 1000 Genomes Project
Shuaishuai Tai
The 1000 Genomes Project
Honglong Wu
The 1000 Genomes Project
Hancheng Zheng
The 1000 Genomes Project
Xiaole Zheng
The 1000 Genomes Project
Yan Zhou
The 1000 Genomes Project
Guoqing Li
The 1000 Genomes Project
Jian Wang
The 1000 Genomes Project
Huanming Yang
The 1000 Genomes Project
Gabor T. Marth
The 1000 Genomes Project
Erik P. Garrison
The 1000 Genomes Project
Weichun Huang
The 1000 Genomes Project
Amit Indap
The 1000 Genomes Project
Deniz Kural
The 1000 Genomes Project
Wan-Ping Lee
The 1000 Genomes Project
Wen Fung Leong
The 1000 Genomes Project
Aaron R. Quinlan
The 1000 Genomes Project
Chip Stewart
The 1000 Genomes Project
Michael P. Stromberg
The 1000 Genomes Project
Alistair N. Ward
The 1000 Genomes Project
Jiantao Wu
The 1000 Genomes Project
Charles Lee
The 1000 Genomes Project
Ryan E. Mills
The 1000 Genomes Project
Xinghua Shi
The 1000 Genomes Project
Mark J. Daly
The 1000 Genomes Project
Mark A. DePristo
The 1000 Genomes Project
David L. Altshuler
The 1000 Genomes Project
Aaron D. Ball
The 1000 Genomes Project
Eric Banks
The 1000 Genomes Project
Toby Bloom
The 1000 Genomes Project
Brian L. Browning
The 1000 Genomes Project
Kristian Cibulskis
The 1000 Genomes Project
Tim J. Fennell
The 1000 Genomes Project
Kiran V. Garimella
The 1000 Genomes Project
Sharon R. Grossman
The 1000 Genomes Project
Robert E. Handsaker
The 1000 Genomes Project
Matt Hanna
The 1000 Genomes Project
Chris Hartl
The 1000 Genomes Project
David B. Jaffe
The 1000 Genomes Project
Andrew M. Kernytsky
The 1000 Genomes Project
Joshua M. Korn
The 1000 Genomes Project
Heng Li
The 1000 Genomes Project
Jared R. Maguire
The 1000 Genomes Project
Steven A. McCarroll
The 1000 Genomes Project
Aaron McKenna
The 1000 Genomes Project
James C. Nemesh
The 1000 Genomes Project
Anthony A. Philippakis
The 1000 Genomes Project
Ryan E. Poplin
The 1000 Genomes Project
Alkes Price
The 1000 Genomes Project
Manuel A. Rivas
The 1000 Genomes Project
Pardis C. Sabeti
The 1000 Genomes Project
Stephen F. Schaffner
The 1000 Genomes Project
Erica Shefler
The 1000 Genomes Project
Ilya A. Shlyakhter
The 1000 Genomes Project
David N. Cooper
The 1000 Genomes Project
Edward V. Ball
The 1000 Genomes Project
Matthew Mort
The 1000 Genomes Project
Andrew D. Phillips
The 1000 Genomes Project
Peter D. Stenson
The 1000 Genomes Project
Jonathan Sebat
The 1000 Genomes Project
Vladimir Makarov
The 1000 Genomes Project
Kenny Ye
The 1000 Genomes Project
Seungtai C. Yoon
The 1000 Genomes Project
Carlos D. Bustamante
The 1000 Genomes Project
Andrew G. Clark
The 1000 Genomes Project
Adam Boyko
The 1000 Genomes Project
Jeremiah Degenhardt
The 1000 Genomes Project
Simon Gravel
The 1000 Genomes Project
Ryan N. Gutenkunst
The 1000 Genomes Project
Mark Kaganovich
The 1000 Genomes Project
Alon Keinan
The 1000 Genomes Project
Phil Lacroute
The 1000 Genomes Project
Xin Ma
The 1000 Genomes Project
Andy Reynolds
The 1000 Genomes Project
Laura Clarke
The 1000 Genomes Project
Paul Flicek
The 1000 Genomes Project
Fiona Cunningham
The 1000 Genomes Project
Javier Herrero
The 1000 Genomes Project
Stephen Keenen
The 1000 Genomes Project
Eugene Kulesha
The 1000 Genomes Project
Rasko Leinonen
The 1000 Genomes Project
William M. McLaren
The 1000 Genomes Project
Rajesh Radhakrishnan
The 1000 Genomes Project
Richard E. Smith
The 1000 Genomes Project
Vadim Zalunin
The 1000 Genomes Project
Xiangqun Zheng-Bradley
The 1000 Genomes Project
Jan O. Korbel
The 1000 Genomes Project
Adrian M. Stütz
The 1000 Genomes Project
Sean Humphray
The 1000 Genomes Project
Markus Bauer
The 1000 Genomes Project
R. Keira Cheetham
The 1000 Genomes Project
Tony Cox
The 1000 Genomes Project
Michael Eberle
The 1000 Genomes Project
Terena James
The 1000 Genomes Project
Scott Kahn
The 1000 Genomes Project
Lisa Murray
The 1000 Genomes Project
Aravinda Chakravarti
The 1000 Genomes Project
Kai Ye
The 1000 Genomes Project
Francisco M. De La Vega
The 1000 Genomes Project
Yutao Fu
The 1000 Genomes Project
Fiona C. L. Hyland
The 1000 Genomes Project
Jonathan M. Manning
The 1000 Genomes Project
Stephen F. McLaughlin
The 1000 Genomes Project
Heather E. Peckham
The 1000 Genomes Project
Onur Sakarya
The 1000 Genomes Project
Yongming A. Sun
The 1000 Genomes Project
Eric F. Tsung
The 1000 Genomes Project
Mark A. Batzer
The 1000 Genomes Project
Miriam K. Konkel
The 1000 Genomes Project
Jerilyn A. Walker
The 1000 Genomes Project
Ralf Sudbrak
The 1000 Genomes Project
Marcus W. Albrecht
The 1000 Genomes Project
Vyacheslav S. Amstislavskiy
The 1000 Genomes Project
Ralf Herwig
The 1000 Genomes Project
Dimitri V. Parkhomchuk
The 1000 Genomes Project
Stephen T. Sherry
The 1000 Genomes Project
Richa Agarwala
The 1000 Genomes Project
Hoda M. Khouri
The 1000 Genomes Project
Aleksandr O. Morgulis
The 1000 Genomes Project
Justin E. Paschall
The 1000 Genomes Project
Lon D. Phan
The 1000 Genomes Project
Kirill E. Rotmistrovsky
The 1000 Genomes Project
Robert D. Sanders
The 1000 Genomes Project
Martin F. Shumway
The 1000 Genomes Project
Chunlin Xiao
The 1000 Genomes Project
Gil A. McVean
The 1000 Genomes Project
Adam Auton
The 1000 Genomes Project
Zamin Iqbal
The 1000 Genomes Project
Gerton Lunter
The 1000 Genomes Project
Jonathan L. Marchini
The 1000 Genomes Project
Loukas Moutsianas
The 1000 Genomes Project
Simon Myers
The 1000 Genomes Project
Afidalina Tumian
The 1000 Genomes Project
Brian Desany
The 1000 Genomes Project
James Knight
The 1000 Genomes Project
Roger Winer
The 1000 Genomes Project
David W. Craig
The 1000 Genomes Project
Steve M. Beckstrom-Sternberg
The 1000 Genomes Project
Alexis Christoforides
The 1000 Genomes Project
Ahmet A. Kurdoglu
The 1000 Genomes Project
John V. Pearson
The 1000 Genomes Project
Shripad A. Sinari
The 1000 Genomes Project
Waibhav D. Tembe
The 1000 Genomes Project
David Haussler
The 1000 Genomes Project
Angie S. Hinrichs
The 1000 Genomes Project
Sol J. Katzman
The 1000 Genomes Project
Andrew Kern
The 1000 Genomes Project
Robert M. Kuhn
The 1000 Genomes Project
Molly Przeworski
The 1000 Genomes Project
Ryan D. Hernandez
The 1000 Genomes Project
Bryan Howie
The 1000 Genomes Project
Joanna L. Kelley
The 1000 Genomes Project
S. Cord Melton
The 1000 Genomes Project
Gonçalo R. Abecasis
The 1000 Genomes Project
Yun Li
The 1000 Genomes Project
Paul Anderson
The 1000 Genomes Project
Tom Blackwell
The 1000 Genomes Project
Wei Chen
The 1000 Genomes Project
William O. Cookson
The 1000 Genomes Project
Jun Ding
The 1000 Genomes Project
Hyun Min Kang
The 1000 Genomes Project
Mark Lathrop
The 1000 Genomes Project
Liming Liang
The 1000 Genomes Project
Miriam F. Moffatt
The 1000 Genomes Project
Paul Scheet
The 1000 Genomes Project
Carlo Sidore
The 1000 Genomes Project
Matthew Snyder
The 1000 Genomes Project
Xiaowei Zhan
The 1000 Genomes Project
Sebastian Zöllner
The 1000 Genomes Project
Philip Awadalla
The 1000 Genomes Project
Ferran Casals
The 1000 Genomes Project
Youssef Idaghdour
The 1000 Genomes Project
John Keebler
The 1000 Genomes Project
Eric A. Stone
The 1000 Genomes Project
Martine Zilversmit
The 1000 Genomes Project
Lynn Jorde
The 1000 Genomes Project
Jinchuan Xing
The 1000 Genomes Project
Evan E. Eichler
The 1000 Genomes Project
Gozde Aksay
The 1000 Genomes Project
Can Alkan
The 1000 Genomes Project
Iman Hajirasouliha
The 1000 Genomes Project
Fereydoun Hormozdiari
The 1000 Genomes Project
Jeffrey M. Kidd
The 1000 Genomes Project
S. Cenk Sahinalp
The 1000 Genomes Project
Peter H. Sudmant
The 1000 Genomes Project
Elaine R. Mardis
The 1000 Genomes Project
Ken Chen
The 1000 Genomes Project
Asif Chinwalla
The 1000 Genomes Project
Li Ding
The 1000 Genomes Project
Daniel C. Koboldt
The 1000 Genomes Project
Mike D. McLellan
The 1000 Genomes Project
David Dooling
The 1000 Genomes Project
George Weinstock
The 1000 Genomes Project
John W. Wallis
The 1000 Genomes Project
Michael C. Wendl
The 1000 Genomes Project
Qunyuan Zhang
The 1000 Genomes Project
Richard M. Durbin
The 1000 Genomes Project
Cornelis A. Albers
The 1000 Genomes Project
Qasim Ayub
The 1000 Genomes Project
Senduran Balasubramaniam
The 1000 Genomes Project
Jeffrey C. Barrett
The 1000 Genomes Project
David M. Carter
The 1000 Genomes Project
Yuan Chen
The 1000 Genomes Project
Donald F. Conrad
The 1000 Genomes Project
Petr Danecek
The 1000 Genomes Project
Emmanouil T. Dermitzakis
The 1000 Genomes Project
Min Hu
The 1000 Genomes Project
Ni Huang
The 1000 Genomes Project
Matt E. Hurles
The 1000 Genomes Project
Hanjun Jin
The 1000 Genomes Project
Luke Jostins
The 1000 Genomes Project
Thomas M. Keane
The 1000 Genomes Project
Si Quang Le
The 1000 Genomes Project
Sarah Lindsay
The 1000 Genomes Project
Quan Long
The 1000 Genomes Project
Daniel G. MacArthur
The 1000 Genomes Project
Stephen B. Montgomery
The 1000 Genomes Project
Leopold Parts
The 1000 Genomes Project
James Stalker
The 1000 Genomes Project
Chris Tyler-Smith
The 1000 Genomes Project
Klaudia Walter
The 1000 Genomes Project
Yujun Zhang
The 1000 Genomes Project
Mark B. Gerstein
The 1000 Genomes Project
Michael Snyder
The 1000 Genomes Project
Alexej Abyzov
The 1000 Genomes Project
Suganthi Balasubramanian
The 1000 Genomes Project
Robert Bjornson
The 1000 Genomes Project
Jiang Du
The 1000 Genomes Project
Fabian Grubert
The 1000 Genomes Project
Lukas Habegger
The 1000 Genomes Project
Rajini Haraksingh
The 1000 Genomes Project
Justin Jee
The 1000 Genomes Project
Ekta Khurana
The 1000 Genomes Project
Hugo Y. K. Lam
The 1000 Genomes Project
Jing Leng
The 1000 Genomes Project
Xinmeng Jasmine Mu
The 1000 Genomes Project
Alexander E. Urban
The 1000 Genomes Project
Zhengdong Zhang
The 1000 Genomes Project
Yingrui Li
The 1000 Genomes Project
Ruibang Luo
The 1000 Genomes Project
Gabor T. Marth
The 1000 Genomes Project
Erik P. Garrison
The 1000 Genomes Project
Deniz Kural
The 1000 Genomes Project
Aaron R. Quinlan
The 1000 Genomes Project
Chip Stewart
The 1000 Genomes Project
Michael P. Stromberg
The 1000 Genomes Project
Alistair N. Ward
The 1000 Genomes Project
Jiantao Wu
The 1000 Genomes Project
Charles Lee
The 1000 Genomes Project
Ryan E. Mills
The 1000 Genomes Project
Xinghua Shi
The 1000 Genomes Project
Steven A. McCarroll
The 1000 Genomes Project
Eric Banks
The 1000 Genomes Project
Mark A. DePristo
The 1000 Genomes Project
Robert E. Handsaker
The 1000 Genomes Project
Chris Hartl
The 1000 Genomes Project
Joshua M. Korn
The 1000 Genomes Project
Heng Li
The 1000 Genomes Project
James C. Nemesh
The 1000 Genomes Project
Jonathan Sebat
The 1000 Genomes Project
Vladimir Makarov
The 1000 Genomes Project
Kenny Ye
The 1000 Genomes Project
Seungtai C. Yoon
The 1000 Genomes Project
Jeremiah Degenhardt
The 1000 Genomes Project
Mark Kaganovich
The 1000 Genomes Project
Laura Clarke
The 1000 Genomes Project
Richard E. Smith
The 1000 Genomes Project
Xiangqun Zheng-Bradley
The 1000 Genomes Project
Jan O. Korbel
The 1000 Genomes Project
Sean Humphray
The 1000 Genomes Project
R. Keira Cheetham
The 1000 Genomes Project
Michael Eberle
The 1000 Genomes Project
Scott Kahn
The 1000 Genomes Project
Lisa Murray
The 1000 Genomes Project
Kai Ye
The 1000 Genomes Project
Francisco M. De La Vega
The 1000 Genomes Project
Yutao Fu
The 1000 Genomes Project
Heather E. Peckham
The 1000 Genomes Project
Yongming A. Sun
The 1000 Genomes Project
Mark A. Batzer
The 1000 Genomes Project
Miriam K. Konkel
The 1000 Genomes Project
Jerilyn A. Walker
The 1000 Genomes Project
Chunlin Xiao
The 1000 Genomes Project
Zamin Iqbal
The 1000 Genomes Project
Brian Desany
The 1000 Genomes Project
Tom Blackwell
The 1000 Genomes Project
Matthew Snyder
The 1000 Genomes Project
Jinchuan Xing
The 1000 Genomes Project
Evan E. Eichler
The 1000 Genomes Project
Gozde Aksay
The 1000 Genomes Project
Can Alkan
The 1000 Genomes Project
Iman Hajirasouliha
The 1000 Genomes Project
Fereydoun Hormozdiari
The 1000 Genomes Project
Jeffrey M. Kidd
The 1000 Genomes Project
Ken Chen
The 1000 Genomes Project
Asif Chinwalla
The 1000 Genomes Project
Li Ding
The 1000 Genomes Project
Mike D. McLellan
The 1000 Genomes Project
John W. Wallis
The 1000 Genomes Project
Matt E. Hurles
The 1000 Genomes Project
Donald F. Conrad
The 1000 Genomes Project
Klaudia Walter
The 1000 Genomes Project
Yujun Zhang
The 1000 Genomes Project
Mark B. Gerstein
The 1000 Genomes Project
Michael Snyder
The 1000 Genomes Project
Alexej Abyzov
The 1000 Genomes Project
Jiang Du
The 1000 Genomes Project
Fabian Grubert
The 1000 Genomes Project
Rajini Haraksingh
The 1000 Genomes Project
Justin Jee
The 1000 Genomes Project
Ekta Khurana
The 1000 Genomes Project
Hugo Y. K. Lam
The 1000 Genomes Project
Jing Leng
The 1000 Genomes Project
Xinmeng Jasmine Mu
The 1000 Genomes Project
Alexander E. Urban
The 1000 Genomes Project
Zhengdong Zhang
The 1000 Genomes Project
Richard A. Gibbs
The 1000 Genomes Project
Matthew Bainbridge
The 1000 Genomes Project
Danny Challis
The 1000 Genomes Project
Cristian Coafra
The 1000 Genomes Project
Huyen Dinh
The 1000 Genomes Project
Christie Kovar
The 1000 Genomes Project
Sandy Lee
The 1000 Genomes Project
Donna Muzny
The 1000 Genomes Project
Lynne Nazareth
The 1000 Genomes Project
Jeff Reid
The 1000 Genomes Project
Aniko Sabo
The 1000 Genomes Project
Fuli Yu
The 1000 Genomes Project
Jin Yu
The 1000 Genomes Project
Gabor T. Marth
The 1000 Genomes Project
Erik P. Garrison
The 1000 Genomes Project
Amit Indap
The 1000 Genomes Project
Wen Fung Leong
The 1000 Genomes Project
Aaron R. Quinlan
The 1000 Genomes Project
Chip Stewart
The 1000 Genomes Project
Alistair N. Ward
The 1000 Genomes Project
Jiantao Wu
The 1000 Genomes Project
Kristian Cibulskis
The 1000 Genomes Project
Tim J. Fennell
The 1000 Genomes Project
Stacey B. Gabriel
The 1000 Genomes Project
Kiran V. Garimella
The 1000 Genomes Project
Chris Hartl
The 1000 Genomes Project
Erica Shefler
The 1000 Genomes Project
Carrie L. Sougnez
The 1000 Genomes Project
Jane Wilkinson
The 1000 Genomes Project
Andrew G. Clark
The 1000 Genomes Project
Simon Gravel
The 1000 Genomes Project
Fabian Grubert
The 1000 Genomes Project
Laura Clarke
The 1000 Genomes Project
Paul Flicek
The 1000 Genomes Project
Richard E. Smith
The 1000 Genomes Project
Xiangqun Zheng-Bradley
The 1000 Genomes Project
Stephen T. Sherry
The 1000 Genomes Project
Hoda M. Khouri
The 1000 Genomes Project
Justin E. Paschall
The 1000 Genomes Project
Martin F. Shumway
The 1000 Genomes Project
Chunlin Xiao
The 1000 Genomes Project
Gil A. McVean
The 1000 Genomes Project
Sol J. Katzman
The 1000 Genomes Project
Gonçalo R. Abecasis
The 1000 Genomes Project
Tom Blackwell
The 1000 Genomes Project
Elaine R. Mardis
The 1000 Genomes Project
David Dooling
The 1000 Genomes Project
Lucinda Fulton
The 1000 Genomes Project
Robert Fulton
The 1000 Genomes Project
Daniel C. Koboldt
The 1000 Genomes Project
Richard M. Durbin
The 1000 Genomes Project
Senduran Balasubramaniam
The 1000 Genomes Project
Allison Coffey
The 1000 Genomes Project
Thomas M. Keane
The 1000 Genomes Project
Daniel G. MacArthur
The 1000 Genomes Project
Aarno Palotie
The 1000 Genomes Project
Carol Scott
The 1000 Genomes Project
James Stalker
The 1000 Genomes Project
Chris Tyler-Smith
The 1000 Genomes Project
Mark B. Gerstein
The 1000 Genomes Project
Suganthi Balasubramanian
The 1000 Genomes Project
Aravinda Chakravarti
The 1000 Genomes Project
Bartha M. Knoppers
The 1000 Genomes Project
Gonçalo R. Abecasis
The 1000 Genomes Project
Carlos D. Bustamante
The 1000 Genomes Project
Neda Gharani
The 1000 Genomes Project
Richard A. Gibbs
The 1000 Genomes Project
Lynn Jorde
The 1000 Genomes Project
Jane S. Kaye
The 1000 Genomes Project
Alastair Kent
The 1000 Genomes Project
Taosha Li
The 1000 Genomes Project
Amy L. McGuire
The 1000 Genomes Project
Gil A. McVean
The 1000 Genomes Project
Pilar N. Ossorio
The 1000 Genomes Project
Charles N. Rotimi
The 1000 Genomes Project
Yeyang Su
The 1000 Genomes Project
Lorraine H. Toji
The 1000 Genomes Project
Chris TylerSmith
The 1000 Genomes Project
Lisa D. Brooks
The 1000 Genomes Project
Adam L. Felsenfeld
The 1000 Genomes Project
Jean E. McEwen
The 1000 Genomes Project
Assya Abdallah
The 1000 Genomes Project
Christopher R. Juenger
The 1000 Genomes Project
Nicholas C. Clemm
The 1000 Genomes Project
Francis S. Collins
The 1000 Genomes Project
Audrey Duncanson
The 1000 Genomes Project
Eric D. Green
The 1000 Genomes Project
Mark S. Guyer
The 1000 Genomes Project
Jane L. Peterson
The 1000 Genomes Project
Alan J. Schafer
The 1000 Genomes Project
Gonçalo R. Abecasis
The 1000 Genomes Project
David L. Altshuler
The 1000 Genomes Project
Adam Auton
The 1000 Genomes Project
Lisa D. Brooks
The 1000 Genomes Project
Richard M. Durbin
The 1000 Genomes Project
Richard A. Gibbs
The 1000 Genomes Project
Matt E. Hurles
The 1000 Genomes Project
Gil A. McVean
The 1000 Genomes Project

Notes

1
To whom correspondence should be addressed. E-mail: [email protected].
Author contributions: S.G., B.M.H., G.T.M., A.G.C., F.Y., R.A.G., 1000G, and C.D.B. designed research; S.G., A.R.I., G.T.M., F.Y., R.A.G., and 1000G performed research; S.G., R.N.G., and A.R.I. contributed new reagents/analytic tools; S.G. analyzed data; and S.G., B.M.H., and C.D.B. wrote the paper.

Competing Interests

The authors declare no conflict of interest.

Metrics & Citations

Metrics

Note: The article usage is presented with a three- to four-day delay and will update daily once available. Due to ths delay, usage data will not appear immediately following publication. Citation information is sourced from Crossref Cited-by service.


Citation statements




Altmetrics

Citations

If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download.

Cited by

    Loading...

    View Options

    View options

    PDF format

    Download this article as a PDF file

    DOWNLOAD PDF

    Get Access

    Login options

    Check if you have access through your login credentials or your institution to get full access on this article.

    Personal login Institutional Login

    Recommend to a librarian

    Recommend PNAS to a Librarian

    Purchase options

    Purchase this article to access the full text.

    Single Article Purchase

    Demographic history and rare allele sharing among human populations
    Proceedings of the National Academy of Sciences
    • Vol. 108
    • No. 29
    • pp. 11727-12185

    Media

    Figures

    Tables

    Other

    Share

    Share

    Share article link

    Share on social media