Clinal distribution of human genomic diversity across the Netherlands despite archaeological evidence for genetic discontinuities in Dutch population history
© Lao et al.; licensee BioMed Central Ltd. 2013
Received: 6 September 2012
Accepted: 26 March 2013
Published: 20 May 2013
The presence of a southeast to northwest gradient across Europe in human genetic diversity is a well-established observation and has recently been confirmed by genome-wide single nucleotide polymorphism (SNP) data. This pattern is traditionally explained by major prehistoric human migration events in Palaeolithic and Neolithic times. Here, we investigate whether (similar) spatial patterns in human genomic diversity also occur on a micro-geographic scale within Europe, such as in the Netherlands, and if so, whether these patterns could also be explained by more recent demographic events, such as those that occurred in Dutch population history.
We newly collected data on a total of 999 Dutch individuals sampled at 54 sites across the country at 443,816 autosomal SNPs using the Genome-Wide Human SNP Array 5.0 (Affymetrix). We studied the individual genetic relationships by means of classical multidimensional scaling (MDS) using different genetic distance matrices, spatial ancestry analysis (SPA), and ADMIXTURE software. We further performed dedicated analyses to search for spatial patterns in the genomic variation and conducted simulations (SPLATCHE2) to provide a historical interpretation of the observed spatial patterns.
We detected a subtle but clearly noticeable genomic population substructure in the Dutch population, allowing differentiation of a north-eastern, central-western, central-northern and a southern group. Furthermore, we observed a statistically significant southeast to northwest cline in the distribution of genomic diversity across the Netherlands, similar to earlier findings from across Europe. Simulation analyses indicate that this genomic gradient could similarly be caused by ancient as well as by the more recent events in Dutch history.
Considering the strong archaeological evidence for genetic discontinuity in the Netherlands, we interpret the observed clinal pattern of genomic diversity as being caused by recent rather than ancient events in Dutch population history. We therefore suggest that future human population genetic studies pay more attention to recent demographic history in interpreting genetic clines. Furthermore, our study demonstrates that genetic population substructure is detectable on a small geographic scale in Europe despite recent demographic events, a finding we consider potentially relevant for future epidemiological and forensic studies.
KeywordsPopulation substructure Genetic cline Genome-wide diversity SNP Europe Netherlands
The presence of genetic gradients across Europe has been described and discussed for more than 30 years. In the case of autosomal markers, a southeast to northwest gradual change in the distribution of the genetic diversity has been reported using principal component analysis (PCA) [1, 2]. Initially, this gradient was described from classical markers such as blood groups , and later was confirmed by genome-wide single nucleotide polymorphisms (SNPs) [3, 4]. This genetic diversity cline is traditionally explained by several major prehistoric demographic events in Europe: the first colonization of Europe by anatomically modern humans together with a postglacial re-expansion from the southern European refugee areas in Palaeolithic times, and the introduction of the Neolithic agricultural lifestyle by people from the Near East . Theoretical studies using computer simulations  have shown that such major prehistoric demographic events can produce genetic gradients in autosomal markers that in particular situations resemble what is observed in real data from Europe. However, simulations tend to necessarily simplify the demographic history by ignoring more subtle demographic events that took place throughout history at a smaller geographical scale such as those in Europe . Furthermore, it was suggested that caution should be taken when interpreting results from PCA analyses . With this study we aim to investigate whether (similar) spatial patterns in genomic diversity can also be detected on a micro-geographic scale, within a European country like the Netherlands, and if so, whether these patterns could also be explained by more recent demographic events.
Given the archaeological, geological and historical evidence for genetic discontinuity in Dutch population history, one might expect that the ancient genetic signatures from Palaeolithic and Neolithic times, such as the southeast to northwest cline seen across Europe, would not be detectable in the contemporary Dutch gene pool. To test this hypothesis via studying the spatial distribution of the Dutch genomic diversity, including computer simulations, and to investigate the overall genomic-geographic substructure of the Dutch population, we sampled 999 individuals at 54 sites across the Netherlands following a grid-like scheme. DNA samples were genotyped with the Genome-Wide Human SNP Array 5.0 (Affymetrix; http://www.affymetrix.com/estore/) from which 443,816 quality-controlled genome-wide autosomal SNPs were used in various spatial, cluster, and simulation analyses.
Dutch subpopulations studied, their sampling coordinates, and sample size before and after data cleaning a
Bergen op Zoom
Each individual was genotyped with the GeneChip Human Mapping 500 K Array Set (Affymetrix) and genotypes were inferred with the Bayesian Robust Linear Model with Mahalanobis distance classifier (BRLMM) algorithm. Sampling sites were not considered in the microarray genotyping procedure to avoid batch effects. Individual data cleaning was performed using the Tukey’s approach as applied in Lao et al. . In brief, for each pair of individuals, an identical-by-state (IBS) distance was computed. Within each subpopulation, Tukey’s outlier criterion was applied and individuals either showing large distances to the rest of the individuals of the same subpopulation (genetic outliers), or individuals of pairs showing smaller distances than the observed in all the pairs of the same subpopulation (strongly genetically related), were excluded. It must be noticed that, due to limited sample size in each subpopulation, the power to detect individual genetic outliers can be small. Using this approach, 30 individuals did not pass the quality control and were excluded. SNPs with more than 10% of missing genotypes in at least one subpopulation were also excluded. Hardy-Weinberg Equilibrium (HWE) was tested in all the autosomal SNPs for each subpopulation. SNPs that did not pass HWE in at least one subpopulation after multiple testing were excluded. Of the 443,816 markers, 414,633 autosomal SNPs were considered clean after applying this filter. None of the considered individuals showed a percentage of missing genotypes >2% and therefore there was no further individual exclusion. We next pruned for linkage disequilibrium (LD) by means of ascertaining markers that showed low LD at a distance <500 kb. We computed Kendall’s Tau B statistic  using the contingency table computed between the genotypes of two loci at a distance <500 kb. We included new loci in the final dataset if the absolute value of the statistic was smaller than 0.5 with the ones already included. After LD ascertainment, the number of autosomal markers was 137,662. This set of markers and 969 individuals were used in further analyses, except in the case of the spatial ancestry analysis (SPA) ), where 952 individuals and all the (non-LD pruned) SNPs were used. Data are available for nonprofit research via an institutional website .
A distance measure was obtained by setting 1-Wij for all the pairs. The distance matrix was plotted by means of MDS after adding a constant to the matrix in order to make all the eigenvalues positive . The mean of the first two dimensions by population were compared with the geographic coordinates of the sampling sites by means of a procrustes analysis  as implemented in the protest method of the vegan R package.
Proportions of ancestry for each individual were computed using ADMIXTURE  and FRAPPE , setting the number of groups (K) to 1 to 6. A pie chart map was constructed for K = 5 on ADMIXTURE consensus results (out of 10 independent replicates merged with CLUMPP  using the greedy algorithm implemented in the software) using MapViewer software . CLUMPP  was used to perform a comparison of the outcome of the two clustering algorithms.
where G .s is the not null genotype (taking values 0 for AA,1 for AB and 2 for BB ) of individual at snp s and n is the total number of SNPs for which either individual i and individual j do not contain null genotypes.
in formula 13 of .
The covariance matrix was used to perform a genetic based spatial autocorrelation analysis . We considered 24 distance classes. Overall significance of the autocorrelogram was tested by means of shuffling the individuals at random between the subpopulations and computing the r value for each class distance. We applied the method described by  to propose a combined P value of the autocorrelogram.
where i is the marker of interest, n is the number of markers that are within a distance <50 kb of the marker of interest, Z i is the computed SPA value of the marker i and w ij is the weight between marker i and j (1 if the marker is within the window of 50 kb, otherwise 0). Local Moran’s I statistic takes positive values (indicating positive local autocorrelation) if the value of one SNP is extreme compared to the rest of the genome and it is surrounded by SNPs with values of similar magnitude. A P value was computed by reshuffling the value of the score 1,000 times at random, then computing local Moran’s I statistic for each marker and comparing it with the observed one. A Manhattan plot of the Local Moran’s I statistic value for these markers with a P value <5e-04 was computed using mhtplot function of the gap R package .
We computed Weir and Cockerham’s combined Fst  between pairs of subpopulations with more than 10 individuals (comprising 46 populations). Negative Fst values between pairs of subpopulations were set to 0. Classical multidimensional scaling was performed with this matrix after adding a constant  to prevent negative eigenvalues. Procrustes  was used to compare geographic coordinates with the first two dimensions. Dependence of the genetic distance matrix and geographic distance was assessed by means of Pearson’s correlation and the statistical significance by means of a Mantel test  as implemented in PASSAGE software  using 1,000 iterations. The presence of a geographic gradient in the Fst matrix was tested by means of a Bearing correlogram  using PASSAGE software .
We performed two different SPLATCHE2  simulations in order to assess the impact of genetic discontinuity on the genetic landscape of the Netherlands. We used a map of Europe of 188 columns by 132 rows, sampled 39 cells from the region comprising the Netherlands and simulated 1,000 SNPs of MAF 0.03 in the Netherlands. Both simulations considered the demographic scenario described in Francois et al.  in Europe; that is, the settlement of Europe started 1600 generations ago from the Middle East by hunter-gatherers, and 400 generations ago a second expansion representing the Neolithic took place in the southeast of Europe. Carrying capacity of each cell populated by hunter-gatherers was set to 500 (50 in  with a cell area 9.23 times smaller) and of Neolithic by 5000 (500 in ). Migration rates were set to 0.4 for Palaeolithic and 0.8 for Neolithic and the growth rates to 0.5 in Palaeolithic and 0.8 in Neolithic in order to ensure the full peopling of the European continent. In the second simulation, in addition to these demographic events, we added a genetic discontinuity in the Netherlands, setting the carrying capacity of the coastal cells (representing 28 out of the 39 cells) to 0 between 70 and 35 generations ago, where they were repopulated by migrants from the neighboring populations. A distance matrix between pairs of populations based on Fst was then computed for each simulated model using Arlequin 3.1 , and negative Fst values set to 0. MDS analyses on each distance matrix and comparison of the MDS result with geographic coordinates of the cells was performed by means of a Procrustes analysis. A Bearing correlogram using each Fst distance matrix and geographic coordinates was conducted with PASSAGE software.
Results and discussion
The locations of the 54 Dutch geographic subpopulations from which the 999 individuals were sampled are shown in Figure 2 and further explained in Table 1. As evident, most of the current Dutch territory was sampled evenly with an average sample size across subpopulations of 18 individuals (range between 1 and 65). Overall, about half of the genome-wide autosomal SNPs (53.75%) had a Weir and Cockerham’s Fst value  of zero (or smaller). The mean Fst value across all genome-wide autosomal SNPs used was only 0.003 (after setting negative values to zero) and the mean combined Fst value between pairs of subpopulations was even smaller at 0.00038. These results together demonstrate a very small overall genetic differentiation among the 54 Dutch subpopulations sampled across the entire country. In fact, genetic differentiation between geographic subpopulations from within the Netherlands as observed here is smaller than between geographic subpopulations from within other northern European countries studied thus far in a systematic fashion, such as Sweden . Our genomic results are in agreement with expectations from human populations of small geographic areas, and suggest the absence of strong genetic barriers within the contemporary Dutch population (in addition to the nonexistence of strong geographic barriers). To investigate the spatial distribution of the Dutch genomic diversity as well as the genetic-geographic substructure of the Dutch population, we applied a combination of well-established and recently introduced approaches to the genomic data after stringent quality control on markers and individuals (see Methods for details on quality control).
We additionally explored whether geographically restricted dialects of the Dutch language, which also show north–south gradients as reported elsewhere , could be associated with the genomic diversity pattern we observed across the country. We estimated the amount of genetic variation explained by classifying the 54 subpopulations according to the 6 main dialects (Frisian, Groningen, Overijssel, Southwest Limburg, Brabant and Central Dutch varieties) that were previously identified in a dendrogram analysis by Heeringa . Analysis of Molecular Variance (AMOVA) showed that classifying the Dutch subpopulations by dialect explains a small and statistically nonsignificant proportion of only 0.2% (P(random value >observed value) = 0.99707 after 1,000 iterations) of the total genetic variance observed. This result indicates that dialects are unlikely to have influenced our genomic findings including the spatial distribution of genomic diversity across the Netherlands.
The genome-wide southeast to northwest cline in the distribution of the genomic diversity across the Netherlands observed here via different analyses could be interpreted as fitting the southeast to northwest genetic cline previously found for the whole of Europe [3, 4]. Without any prior knowledge about the geological and human settlement history of the sampled region, one may explain the observed genomic gradient across the Netherlands by the major prehistoric demographic events that were previously used to explain the cline seen across the whole of Europe . However, taking into account the strong palaeogeographic and archaeological evidence for marked population discontinuities on the Dutch territory during several, including more recent, periods in the Dutch history, we regard it as rather unlikely that the Palaeolithic colonization together with postglacial re-colonization and the Neolithic transformation process are directly responsible for the genomic findings we obtained here for the Dutch population. To test if the observed genomic cline could also be explained by recent events in the Dutch history [see Additional file 1 for details], we ran two SPLATCHE2  simulations. In the first simulation, we used the parameters of the Palaeolithic-Neolithic model previously proposed by Francois et al.  (see Methods for details). In the second simulation, we introduced a genetic discontinuity scenario around 250 Anno Domini (AD) (70 generations ago, assuming 25 years per generation) in the Netherlands, when most of the country close to the sea remained uninhabitable by humans (Figure 1) up to 35 generations ago, or until approximately 1250 AD. After this period, previously uninhabitable areas acquired the same carrying capacity as the rest of Europe and became populated by individuals from the surrounding populations in this model. For each simulation, we generated 1,000 SNPs at a minimum allele frequency (MAF) of 0.03 and computed the Fst distance between pairs of populations using Arlequin 3.1 , setting all negative Fst values to 0. The Fst matrix of each of the two demographic models was then used in a MDS analysis and compared by means of Procrustes analysis either with the geographic coordinates or with the MDS coordinates of the other model. We found that both models strongly correlate with geography (correlation with geography in a symmetric Procrustes rotation when using the genetic discontinuity model: 0.576, P value = 0.001; correlation in a symmetric Procrustes rotation of the Palaeolithic-Neolithic model: 0.62, P value = 0.001; both analyses based on 1,000 permutations). Furthermore, we observed that the outcomes of both model simulations are statistically significant in their correlation with each other (correlation in a symmetric Procrustes rotation: 0.446, P value = 0.003). The Bearing correlogram analysis using the Fst distance matrix obtained with the model of genetic discontinuity is highly similar to the one produced by considering genetic continuity (Adjusted R-squared: 0.93, P value <2.2e-16), which suggests that the genetic gradient produced by both models is virtually indistinguishable. This finding, together with the rich archaeological evidence for human genetic discontinuity on Dutch territory led us to propose that the observed genomic gradient across the Netherlands was not caused by ancient but rather by recent events in Dutch history.
Although it cannot be excluded that the observed genomic gradient across the Netherlands that we explain by recent events, by chance resembles the ancient genomic gradient seen across Europe, another explanation is that this gradient was re-introduced by immigration of people from outside regions carrying ancient genetic signatures. One prerequisite for this scenario would be that immigration did not occur by one major population (or a limited number of populations), described as elite-dominance, but by movements of several populations from adjacent areas of similar latitudes in a way that the northern parts of the Netherlands received immigrants from northern/northeastern neighboring regions, southern parts from southern/southeastern neighboring regions, and central parts from eastern neighboring regions. Also, the mainly south–north geographic orientation of the Dutch territory provides a suitable prerequisite for such a scenario given the south–north genomic cline observed across Europe. However, there is no clear evidence provided by the archaeological records that would support such a scenario. The observation that subpopulations from the central-east of the Netherlands appeared more diverse (within and between groups) on the genome-wide level compared to all other Dutch subpopulations tested, could be indicative of recent admixture with other genetically diverse subpopulations not analyzed in our study. It would require, however, more detailed archaeological and/or historical research in addition to similarly detailed genetic information from regions outside the current Dutch political borders to disentangle the exact demographic events that shaped the current genetic variation of the Dutch population.
Besides evolutionary implications, our findings of small but detectable genomic substructure in the Dutch population, particularly the detection of geographic groups of Dutch subpopulations that can be differentiated using genome-wide data, also is of relevance for epidemiology and forensics. For future epidemiological studies, this knowledge may be relevant for (disease) gene mapping on Dutch individuals to avoid confounding effects that in principle can reveal false-positive findings. For future forensic genetic studies, the implications are two-fold. First, the detected population substructure may be considered as a correction factor when estimating match probabilities of STR profiles obtained from crime scene and suspect materials in the Netherlands. Second, our data provide evidence that in case the large number of SNPs used here can be derived from a forensic DNA sample, inferring the subregion of biogeographic ancestry within the Netherlands of an unknown may be feasible, which can provide useful investigative information to find unknown perpetrators.
We have shown that despite the genetic differentiation between Dutch individuals and subpopulations sampled systematically across the country being very small, the overall genome-wide diversity tends to correlate statistically significantly with geography and that the genomic map of the Netherlands resembles the geographic map of sampling locations in all dedicated analyses we performed. Furthermore, we identified a significant southeast to northwest cline in the distribution of genomic diversity across the Netherlands, similar to earlier findings from across Europe. For the Netherlands however, the classical interpretation of the observed genetic gradient by Paleolithic-Neolithic processes is challenged by the geological, archaeological and historical evidence pointing towards population discontinuity on the Dutch territory through the ages. Our demographic simulations revealed that the expected Paleolithic-Neolithic pattern in autochthonous populations would be similar to the one produced by a recent colonization of a region from neighboring areas. Considering the evidence for population discontinuity we therefore believe that the genomic patterns we observe are caused by recent rather than ancient events in the Dutch population history. On a wider picture, our results indicate that local and more recent demographic events can produce genetic patterns strongly resembling those traditionally explained by the major prehistoric migrations. We therefore suggest that future studies pay more attention to local and more recent demographic events when explaining clinal distributions of genetic diversity. Ultimately, ancient DNA analysis of past populations in comparison with DNA analysis of contemporary populations from the same region should be used to elucidate the contribution of ancient versus recent populations to the current gene pool of the Netherlands.
Bayesian robust linear model with Mahalanobis distance classifier algorithm
Forensic Laboratory for DNA Research of the Leiden University Medical Center
Minimum allele frequency
Principal component analysis
Single nucleotide polymorphisms
Spatial ancestry analysis.
We thank Miguel Arenas Busto and Stefano Mona for valuable help in running SPLATCHE2. Susan Walsh is acknowledged for valuable comments on the manuscript. This study was supported in part by funding from the Netherlands Forensic Institute (NFI) and by a grant from the Netherlands Genomics Initiative (NGI)/Netherlands Organization for Scientific Research (NWO) within the framework of the Forensic Genomics Consortium Netherlands (FGCN).
- Cavalli-Sforza LL, Menozzi P, Piazza A: The history and geography of human genes. 1994, Princeton (NJ): Princeton University Press
- Menozzi P, Piazza A, Cavalli-Sforza L: Synthetic maps of human gene frequencies in Europeans. Science. 1978, 201: 786-792. 10.1126/science.356262.View ArticlePubMed
- Lao O, Lu TT, Nothnagel M, Junge O, Freitag-Wolf S, Caliebe A, Balascakova M, Bertranpetit J, Bindoff LA, Comas D, Holmlund G, Kouvatsi A, Macek M, Mollet I, Parson W, Palo J, Ploski R, Sajantila A, Tagliabracci A, Gether U, Werge T, Rivadeneira F, Hofman A, Uitterlinden AG, Gieger C, Wichmann HE, Ruther A, Schreiber S, Becker C, Nurnberg P, Nelson MR, Krawczak M, Kayser M: Correlation between genetic and geographic structure in Europe. Curr Biol. 2008, 18: 1241-1248. 10.1016/j.cub.2008.07.049.View ArticlePubMed
- Novembre J, Johnson T, Bryc K, Kutalik Z, Boyko AR, Auton A, Indap A, King KS, Bergmann S, Nelson MR, Stephens M, Bustamante CD: Genes mirror geography within Europe. Nature. 2008, 456: 98-101. 10.1038/nature07331.PubMed CentralView ArticlePubMed
- Francois O, Currat M, Ray N, Han E, Excoffier L, Novembre J: Principal component analysis under population genetic models of range expansion and admixture. Mol Biol Evol. 2010, 27: 1257-1268. 10.1093/molbev/msq010.View ArticlePubMed
- Sokal RR, Oden NL, Walker J, Di Giovanni D, Thomson BA: Historical population movements in Europe influence genetic relationships in modern samples. Hum Biol. 1996, 68: 873-898.PubMed
- Pinhasi R, Thomas MG, Hofreiter M, Currat M, Burger J: The genetic history of Europeans. Trends Genet. 2012, 28: 496-505. 10.1016/j.tig.2012.06.006.View ArticlePubMed
- Hughes JD: The Netherlands: Holland against the sea. An environmental history of the world: Humankind’s changing role in the community of life. Volume 2. 2009, London and New York: Taylor & Francis, 2
- Vos PC, Bazelmans J, Weerts HJT, van der Meulen MJ: Atlas van nederland in het holoceen. 2011, Amsterdam: Uitgeverij Bert Bakker
- Kooijmans LP, van den Broeke P, Fokkens H, van Gijn A: Nederland in de prehistorie. 2005, Amsterdam: Uitgeverij Bert BakkerView Article
- van Beek R: Reliëf in tijd en ruimte; interdisciplinair onderzoek naar bewoning en landschap van oost-nederland tussen de vroege prehistorie en middeleeuwen. 2010, Leiden: Sidestone Press
- Dijkstra M: Landschap en bewoning tussen de 3e en 9e eeuw in Zuid-Holland, in het bijzonder de Oude Rijnstreek. 2011, Leiden: Sidestone Press
- Gerrets DA: Op de grens van land en water; Dynamiek van landschap en samenleving in Frisia gedurende de Romeinse tijd en de volksverhuizingstijd. 2010, Groningen: Barkhuis & Groningen University Library
- van der Velde HM: Cananefaten en Friezen aan de monding van de Rijn, ADC rapport 1456. 2008, Amersfoort: ADC
- Noordhoff Atlas producties: De Bosatlas van de geschiedenis van Nederland. 2011, Groningen
- Ekamper P, van der Erf R, van der Gaag N, Henkens K, van Imhoff E, van Poppel F: Bevolkingsatlas van Nederland; demografische ontwikkelingen van 1850 tot heden. 2003, Rijswijk: Uitgeverij Elmar
- Kendall MG: Rank correlation methods. 1955, New York: Hafner Publishing Co
- Yang WY, Novembre J, Eskin E, Halperin E: A model-based approach for analysis of spatial structure in genetic data. Nat Genet. 2012, 44: 725-731. 10.1038/ng.2285.PubMed CentralView ArticlePubMed
- Forensic molecular biology resources.http://www.erasmusmc.nl/fmb/resources/,
- R Development Core Team: R: A language and environment for statistical computing. 2013, Vienna, Austria: R Foundation for Statistical Computing
- Browning BL, Browning SR: A fast, powerful method for detecting identity by descent. Am J Hum Genet. 2011, 88: 173-182. 10.1016/j.ajhg.2011.01.010.PubMed CentralView ArticlePubMed
- Browning SR, Browning BL: Rapid and accurate haplotype phasing and missing-data inference for whole-genome association studies by use of localized haplotype clustering. Am J Hum Genet. 2007, 81: 1084-1097. 10.1086/521987.PubMed CentralView ArticlePubMed
- Gusev A, Palamara PF, Aponte G, Zhuang Z, Darvasi A, Gregersen P, Pe’er I: The architecture of long-range haplotypes shared within and across populations. Mol Biol Evol. 2012, 29: 473-486. 10.1093/molbev/msr133.PubMed CentralView ArticlePubMed
- Cox TF, Cox MAA: Multidimensional scaling. 2001, Florida: Chapman & Hall/CRC, 2
- Wang C, Szpiech ZA, Degnan JH, Jakobsson M, Pemberton TJ, Hardy JA, Singleton AB, Rosenberg NA: Comparing spatial maps of human population-genetic variation using procrustes analysis. Stat Appl Genet Mol Biol. 2010, 9: Article 13-PubMed CentralPubMed
- Alexander DH, Novembre J, Lange K: Fast model-based estimation of ancestry in unrelated individuals. Genome Res. 2009, 19: 1655-1664. 10.1101/gr.094052.109.PubMed CentralView ArticlePubMed
- Tang H, Peng J, Wang P, Risch NJ: Estimation of individual admixture: analytical and study design considerations. Genet Epidemiol. 2005, 28: 289-301. 10.1002/gepi.20064.View ArticlePubMed
- Jakobsson M, Rosenberg NA: CLUMPP: a cluster matching and permutation program for dealing with label switching and multimodality in analysis of population structure. Bioinformatics. 2007, 23: 1801-1806. 10.1093/bioinformatics/btm233.View ArticlePubMed
- Software G: Mapviewer 7.1. Thematic and Analytical Mapping System. 2006, Golden Software: Colorado
- Smouse PE, Peakall R: Spatial autocorrelation analysis of individual multiallele and multilocus genetic structure. Heredity (Edinb). 1999, 82 (Pt 5): 561-573.View Article
- Wilson DJ, McVean G: Estimating diversifying selection and functional constraint in the presence of recombination. Genetics. 2006, 172: 1411-1425.PubMed CentralView ArticlePubMed
- Schabenberger O, Gotway CA: Statistical methods for spatial data analysis. 2005, Boca Raton: Chapman & Hall/CRC
- Zhao JH: gap: Genetic Analysis Package. J Stat Softw. 2007, 23: 1-18.View Article
- Weir BS, Cockerham CC: Estimating F-statistics for the analysis of population structure. Evolution. 1984, 38: 1358-1370. 10.2307/2408641.View Article
- Mantel N: The detection of disease clustering and a generalized regression approach. Cancer Res. 1967, 27: 209-220.PubMed
- Rosenberg MS, Anderson CD: PASSaGE: Pattern Analysis, Spatial Statistics and Geographic Exegesis. Version 2. Methods Ecol Evol. 2011, 2: 229-232. 10.1111/j.2041-210X.2010.00081.x.View Article
- Rosenberg MS: The bearing correlogram: a new method of analyzing directional spatial autocorrelation. Geogr Anal. 2000, 32: 267-278.View Article
- Ray N, Currat M, Foll M, Excoffier L: SPLATCHE2: a spatially explicit simulation framework for complex demography, genetic admixture and recombination. Bioinformatics. 2010, 26: 2993-2994. 10.1093/bioinformatics/btq579.View ArticlePubMed
- Excoffier L, Laval G, Schneider S: Arlequin ver. 3.0: An integrated software package for population genetics data analysis. Evol Bioinformatics Online. 2005, 1: 47-50.
- Humphreys K, Grankvist A, Leu M, Hall P, Liu J, Ripatti S, Rehnstrom K, Groop L, Klareskog L, Ding B, Grönberg H, Xu J, Pedersen NL, Lichtenstein P, Mattingsdal M, Andreassen OA, O'Dushlaine C, Purcell SM, Sklar P, Sullivan PF, Hultman CM, Palmgren J, Magnusson PK: The genetic structure of the Swedish population. PLoS One. 2011, 6: e22547-10.1371/journal.pone.0022547.PubMed CentralView ArticlePubMed
- Fraley C, Raftery AE: MCLUST Version 3 for R: Normal Mixture Modeling and Model-based Clustering. Technical Report No. 504, Department of Statistics. 2006, University of Washington (revised 2009)
- Kayser M, Liu F, Janssens AC, Rivadeneira F, Lao O, van Duijn K, Vermeulen M, Arp P, Jhamai MM, van Ijcken WF, den Dunnen JT, Heath S, Zelenika D, Despriet DD, Klaver CC, Vingerling JR, de Jong PT, Hofman A, Aulchenko YS, Uitterlinden AG, Oostra BA, van Duijn CM: Three genome-wide association studies and a linkage analysis identify HERC2 as a human iris color gene. Am J Hum Genet. 2008, 82: 411-423. 10.1016/j.ajhg.2007.10.003.PubMed CentralView ArticlePubMed
- Heeringa W: Measuring dialect pronunciation differences using levenshtein distance. 2004, Groningen, Netherlands: University of Groningen, Faculty of Arts, the Humanities Computing department
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.