Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
N Engl J Med. Author manuscript; available in PMC 2012 September 29.
Published in final edited form as:
PMCID: PMC3350841

HLA Class II Locus and Susceptibility to Podoconiosis

Fasil Tekola Ayele, Ph.D., M.P.H., Adebowale Adeyemo, M.D., Chris Finan, Ph.D., Elena Hailu, M.Sc., Paul Sinnott, Ph.D., Natalia Diaz Burlinson, M.Sc., Abraham Aseffa, M.D., Ph.D., Charles N. Rotimi, Ph.D., M.P.H., Melanie J. Newport, M.D., Ph.D., and Gail Davey, M.D.



Podoconiosis is a tropical lymphedema resulting from long-term barefoot exposure to red-clay soil derived from volcanic rock. The World Health Organization recently designated it as a neglected tropical disease. Podoconiosis develops in only a subgroup of exposed people, and studies have shown familial clustering with high heritability (63%).


We conducted a genomewide association study of 194 case patients and 203 controls from southern Ethiopia. Findings were validated by means of family-based association testing in 202 family trios and HLA typing in 94 case patients and 94 controls.


We found a genomewide significant association of podoconiosis with the single-nucleotide polymorphism (SNP) rs17612858, located 5.8 kb from the HLA-DQA1 locus (in the allelic model: odds ratio, 2.44; 95% confidence interval [CI], 1.82 to 3.26; P = 1.42×10−9; and in the additive model: odds ratio, 2.19; 95% CI, 1.66 to 2.90; P = 3.44×10−8), and suggestive associations (P<1.0×10−5) with seven other SNPs in or near HLA-DQB1, HLA-DQA1, and HLA-DRB1. We confirmed these associations using family-based association testing. HLA typing showed the alleles HLA-DRB1*0701 (odds ratio, 2.00), DQA1*0201 (odds ratio, 1.91), and DQB1*0202 (odds ratio, 1.79) and the HLA-DRB1*0701–DQB1*0202 haplotype (odds ratio, 1.92) were risk variants for podoconiosis.


Association between variants in HLA class II loci with podoconiosis (a noncommuni-cable disease) suggests that the condition may be a T-cell–mediated inflammatory disease and is a model for gene–environment interactions that may be relevant to other complex genetic disorders. (Funded by the Wellcome Trust and others.)

Podoconiosis (endemic nonfilarial elephantiasis) is a noninfectious geochemical disease that results in bilateral swelling of the lower legs (Fig. 1). It is found among subsistence farmers whose feet are exposed over many years to red-clay soil derived from volcanic rock. Podoconiosis is an important yet neglected clinical and public health burden in more than 10 countries across tropical Africa, Central and South America, and north India.1 It is estimated that up to 1 million people are affected by podoconiosis in Ethiopia alone.2,3 In the Wolaita zone of southern Ethiopia (where this study was conducted), it affects 1 in 20 people,2 imposes an immense economic burden,4 and causes severe social stigma.5 The World Health Organization recently designated podoconiosis as a neglected tropical disease in recognition of its impact ( Our limited knowledge of its pathogenesis is based on evidence from the 1980s, which suggests that mineral particles in red-clay soils are absorbed through the skin of the foot and engulfed by macrophages in the lower limb lymphatic system, inducing an inflammatory response in the lymphatic vessels, which results in fibrosis and obstruction of the vessel lumen.6 The disease develops in only some people exposed to irritant soils, although mineral particles can be seen in the lymphatic system and lymph nodes of unaffected as well as affected people.7,8 Familial clustering of podoconiosis case patients9 and our pedigree study10 suggest that both genetic and environmental factors are involved in the pathogenesis of podoconiosis. We estimated the heritability of podoconiosis to be 63% and the risk ratio of occurrence among siblings of the case patient to be 5.07, indicating that a sibling of an affected person has a risk of podoconiosis developing that is five times that of a randomly selected person in the general population.10

Figure 1
Podoconiosis (Endemic Nonfilarial Elephantiasis)

Understanding the pathogenesis of podoconiosis will allow for the refinement of current prevention and treatment measures and may lead to important insights into the pathogenesis of other important fibrosing diseases, such as filariasis and pneumoconiosis. Podoconiosis may also represent a simplified model of gene–environment interactions, which remain poorly understood in many complex genetic diseases. We report the findings of a genomewide association study, family-based association testing, and HLA typing in a southern Ethiopian population.


Study Conduct

The study was approved by the institutional ethics review boards of the Medical Faculty of Addis Ababa University and the Armauer Hansen Research Institute–All Africa Leprosy, Tuberculosis and Rehabilitation Training Center as well as the ethics review committee of the Ethiopian Ministry of Science and Technology. The study was conducted according to the protocol, available with the full text of this article at All authors vouch for the accuracy and completeness of the reported data and the fidelity of the study to the protocol. Written informed consent was obtained from all participants following a rapid ethics assessment that explored social, cultural, and economic factors that might influence the consent process for genetic research on podoconiosis. The issues identified in the rapid ethics assessment and the approach followed for obtaining consent for this study have been described elsewhere.11,12 The absence of immediate therapeutic benefit from the genetic study was explained to participants.

Study Participants

Study participants were recruited from the Wolaita zone of southern Ethiopia from podoconiosis-affected families (DNA samples were collected from both parents and two affected siblings); unaffected persons were recruited to serve as controls. All participants were from the same broad geologic area covered by reddish-brown clay soils containing colloid-size particles derived from volcanic basalt rocks.13 The genomewide association study was conducted with the use of samples from one randomly selected affected sibling (of the two recruited) from each family and an unaffected, unrelated control for each case patient. The eligibility criteria for controls were the current absence of podoconiosis, the absence of a personal or family history of podoconiosis, age of at least 50 years (to minimize misclassification of potential case patients as controls), residence in the study area for at least 25 years, exposure to the same irritant clay soil as the case patients, and inconsistent use of shoes. Consequently, the controls were older than the case patients (average age, 62 years vs. 24 years). The family-based association test included the affected sibling who was not included in the genomewide association study as well as both parents from each family.


We obtained a 2-ml saliva sample from all participants. DNA was extracted with the use of a DNA-purification protocol (Oragene, DNA Genotek). DNA was quantified (Quant-iT PicoGreen reagent, Molecular Probes) according to the manufacturer's protocol. Genotyping was performed by a contract company (deCODE Genetics) by means of a chip (HumanHap 610 Bead Chip, Illumina) that contains more than 620,000 single-nucleotide polymorphisms (SNPs). Genotyping for the family-based association testing was performed by KBioscience ( HLA-DRB1, HLA-DQA1, HLA-DQB1, and HLA-DPB1 typing of ap proximately half of the total participants (94 case patients and 94 controls) was done with the use of a high-definition technology (Luminex xMAP) that analyzes DNA samples with sequence-specific oligonucleotides amplified with the use of a polymerase-chain-reaction assay.14 Quality-control and data-analysis procedures are described in the Supplementary Appendix (available at


Enrollment and Samples

Characteristics of the study participants are summarized in Tables 1 and 2 in the Supplementary Appendix. Figure 1 in the Supplementary Appendix summarizes the reasons for exclusion of DNA samples from 8 patients with podoconiosis and 12 controls and of 32,358 SNPs from the genome-wide association analysis. The final data set for the genomewide association analysis comprised 551,840 autosomal SNPs in 194 case patients and 203 controls.

Significant excess sharing of marker alleles identical by descent (from the same ancestral chromosome) owing to cryptic relatedness (genetic relatedness among participants that was not known by the investigator) or sample contamination was ruled out as a reason for exclusion of samples from data analysis because the PI_HAT values (a measure of the degree of genetic relationship between participants) were less than 0.05. Inbreeding coefficients were 0.03 or less for all participants and never exceeded 4 SD of the mean coefficient. Population structure (allele-frequency differences between groups of persons that result in spurious associations) was minimal in the sample because of three factors. First, plots from principal components analysis showed that all case patients belonged to the same cluster, and there was no difference in pairwise identity-by-state distances (degree of marker-allele sharing between pairs of persons) between case patients and controls (P = 0.99). Second, the quantile–quantile plot of the distribution of the test statistic was consistent with the expected distribution under the null hypothesis (Fig. 2 in the Supplementary Appendix). Third, the genomic-control inflation λ value15 was 1.02 in both the allelic and additive models, indicating a lack of population structure. Nonetheless, the first four principal components were included as covariates in the genomewide analysis to account for any potential residual population structure.

Significant Genomewide Association Study Signals

One locus, rs17612858, showed a genomewide significant association with podoconiosis (P = 1.42×10−9 in the allelic model and P = 3.44×10−8 in the additive model). This intergenic SNP lies between HLADQA1 (5.8 kb away) and HLA-DQB1 (6.7 kb away). The presence of the minor allele (T) of this SNP was associated with an increased risk of susceptibility to podoconiosis substantially (odds ratio, 2.44 in the allelic model and 2.19 in the additive model). The 15 SNPs with the most significant association with podoconiosis from the genomewide association analysis are presented in Table 1. Most of these SNPs map to or near genes in the HLA class II locus on chromosome 6: for example, rs9273349 and rs1063355, near HLA-DQB1; rs9270856 and rs9271100, near HLA-DRB1; and rs9271170, rs17843604, and rs17612633, near HLA-DQA1.

Table 1
Genomewide Association Results with the Allelic and Additive Genetic Models.*

Figure 2 shows the plots of P values for the association of each SNP with the risk of podoconiosis and the association signals on chromosome 6. Non-HLA regions with some of the most significant associations included rs2906966 on chromosome 17 (near CDRT4 and FAM18B2), rs11644557 on chromosome 16 (in ITFG1), and rs9514099 on chromosome 13 (near SLC10A2). Genomewide significance was not achieved in the dominant and recessive models, but the most significant association in the recessive model was also significant in the additive and allelic models and had a stronger effect (odds ratio for podoconiosis with rs17612858, 3.40; 95% confidence interval [CI], 2.07 to 5.58) (Tables 3 and 4 in the Supplementary Appendix).

Figure 2 (facing page)
Association of Single-Nucleotide Polymorphisms (SNPs) with Podoconiosis

We estimated the proportion of the variance in the occurrence of podoconiosis that could be explained by the SNP that achieved genomewide significance (rs17612858), as well as all eight SNPs with the most significant associations with the disease, by using Nagelkerke's R2 value in a logistic-regression model involving a backward-stepwise-likelihood ratio method (see the Supplementary Appendix). The eight SNPs explained 15.6% of the variance, and the SNP with the most significant P value (rs17612858) alone explained 9.8% of the variance in occurrence of podoconiosis.

To test for potential interaction between HLA and non-HLA regions, we selected SNPs with P values of less than 10−4 in the additive genetic model and performed epistasis tests between a selection of 15 SNPs in HLA class II loci and a selection of 38 SNPs in non-HLA genes. Of the group of 567 valid tests, only 11 had P values of less than 0.01 (Table 5 in the Supplementary Appendix), and none were significant after Bonferroni correction (with a threshold of P<0.05÷567, or P<8.8×10−5).

A genomewide haplotype test showed that the eight HLA class II SNPs that had the greatest significance in single-marker association analysis formed a haplotype significantly associated with podoconiosis (P = 4.5×10−8) (Table 6 in the Supplementary Appendix). The eight HLA class II SNPs formed one haplotype block, and the average linkage disequilibrium within the block was weak (r2 = 0.41) (Fig. 3 in the Supplementary Appendix). The SNPs were 0.3 to 7 kb from the lead SNP (rs17612858) in the genomewide association study. Conditioning the analysis by including rs17612858 as a covariate made the haplotype association non-significant (P>10−3), implying that rs17612858 explained nearly all of the observed haplotype effect (Table 7 in the Supplementary Appendix). A stepwise multivariate-regression analysis also confirmed that only rs17612858 had an independent effect (odds ratio for having the minor-allele homozygote TT vs. the major-allele homozygote AA, 4.58 [95% CI, 2.61 to 8.04]; odds ratio of TA vs. AA, 1.68 [95% CI, 1.03 to 2.73]).

We found eight SNPs, of which six were heterozygous, within 20 kb of rs17612858 in two HapMap (release 3.2) populations from Africa: Yoruba from Ibadan, Nigeria, and Maasai from Kinyawa, Kenya. Five of the six heterozygous SNPs had minor allele frequencies that differed significantly between the two populations, as well as different haplotype structures (Table 8 and Fig. 4 in the Supplementary Appendix).

Family-Based Association Testing

The family-based association study included 202 family trios. The 24 SNPs with the most significant P values from the genomewide association analysis were selected, of which 21 were successfully genotyped (Table 9 in the Supplementary Appendix). Genotypes at each marker were in Hardy–Weinberg equilibrium (P<0.001 for unrelated individuals). The average genotyping rate was 0.97, and the average marker heterozygosity was 0.40. Single-marker analysis by means of the family-based association test revealed significant associations between podoconiosis and 7 of the 21 SNPs (P<0.05). Alleles of rs1063355, rs17612633, rs17612858, and rs17867526 showed significant overtransmission in both the additive and recessive models. The major alleles of rs17843604 and rs9271100 showed significant over-transmission in the recessive model, and the major allele of rs17211510 showed overtransmission in the additive model. The odds ratio for each of these associations in the additive model was approximately 1.5 (Table 10 in the Supplementary Appendix). Six of the 7 SNPs mapped to the HLA-DQA1, HLA-DQB1, or HLA-DRB1 gene (Table 2).

Table 2
Single-Nucleotide Polymorphism (SNP) Associations with Podoconiosis in Family-Based Association Tests.

Three of the associations maintained statistical significance after applying the Bonferroni correction for a P-value threshold of less than 0.007 in the recessive model. Family-based haplotype analysis showed that the haplotype GCCTTC — formed by the HLA SNPs with significant associations on single-marker family-based association testing — was overtransmitted to the affected siblings (P<0.01) (Table 11 in the Supplementary Appendix).

HLA Class II Associations

We examined the associations between HLA types and podoconiosis by using two models. In the first model, we tested the association with the presence of an allele (vs. its absence). The individual alleles DRB1*0701 (P = 0.02), DQA1*0201 (P = 0.02), and DQB1*0202 (P = 0.03) conferred susceptibility to podoconiosis. There were also significant protective associations with DRB1*0102 (P<0.001), DRB1*1302 (P = 0.04), DQA1*01MV (P<0.001), and DQB1*0501 (P = 0.001). The haplotype constructed from these alleles, DRB1*0102–DQA1*01MV–DQB1*0501, was also associated with podoconiosis (P = 0.01) (Table 3). The full list of HLA alleles and their association statistics is presented in Table 12 in the Supplementary Appendix.

Table 3
Distribution of HLA-DRB1, HLA-DQA1, and HLA-DQB1 Alleles Associated with Podoconiosis.*

In the second model, we tested for association with the presence of two alleles (vs. none) and only one allele (vs. none). In this model, the DRB1*0701–DQB1*0202 haplotype (P = 0.04), as well as either allele (P = 0.02), was associated with an increased odds of having podoconiosis (odds ratio, 1.92) (Table 13 in the Supplementary Appendix).


Using a genomewide approach, we have identified an association between genetic variants in the HLA class II region harboring HLA-DQA1, HLA-DQB1, and HLA-DRB1 and podoconiosis, a neglected tropical disease affecting an estimated 4 million people worldwide. The most strongly associated SNPs were validated in a set of families by means of family-based association testing, and HLA typing showed that specific HLA alleles and haplotypes were significantly associated with differential risk of disease. We estimated that the HLA SNPs associated with podoconiosis in this study explained 15.6% of the genetic variance in podoconiosis, conferring an increase in risk by a factor of 2 to 3.

The association of HLA class II suggests that podoconiosis is a T-cell–mediated inflammatory disease. The initial trigger for T-cell activation is recognition of an antigenic peptide bound to an HLA molecule on antigen-presenting cells.16 Although HLA class II molecules are central to the presentation of exogenous (i.e., foreign and usually pathogen-derived) antigens, they have also been implicated in diseases triggered by minerals, such as silicosis and berylliosis. In berylliosis, which is associated with HLA-DPGlu69, possible mechanisms of pathogenesis include modification of a self peptide by beryllium, allowing for presentation of the self peptide by HLA class II molecules or direct binding of beryllium to, and alteration of, the HLA-peptide binding pocket.17-21 In podoconiosis, the DRB1*0701, DQA1*0201, and DQB1*0202 alleles may have a functional role in antigen presentation to T cells, leading to induction of the immune response and development of disease in response to a currently undefined soil antigen or mineral.

HLA genetic data from African populations are scarce. De Bakker and colleagues included samples from the Nigerian Yoruba population in their construction of a high-resolution HLA and SNP haplotype map.22 This involved typing the classical HLA genes and more than 7500 SNPs that were analyzed together to identify informative tag SNPs that captured the variation in the HLA region. Of the HLA SNPs we found to be associated with podoconiosis, only rs106335 has been identified as an informative tag SNP in the Yoruba. The absence of convergence of more than 1 SNP between the two studies may not be surprising, given the wide diversity of African populations in terms of allele frequencies, local linkage-disequilibrium patterns, and extensive population substructure.23,24 To demonstrate this, we compared the allele frequencies and haplotype structure for 8 SNPs within 20 kb of the most strongly associated marker from our study (rs17612858) among two HapMap (release 3.2) populations from Africa: Yoruba from Ibadan, Nigeria, and Maasai from Kinyawa, Kenya. Six of these 8 SNPs were heterozygous; 5 of these 6 had minor allele frequencies that differed significantly between the two populations, as well as different haplotype structures (Table 8 and Fig. 4 in the Supplementary Appendix). This finding indicates that a single group in Africa cannot be taken as a proxy for predicting HLA alleles of all other populations on the continent, including the Ethiopian Wolaita population.

We selected barefoot, older participants as “supercontrols” — people who had been sufficiently exposed to the environmental risk factor to have podoconiosis develop yet it did not. These controls were included under the assumption that they would possess fewer genetic susceptibility variants than the case patients. Although the use of such controls may introduce bias with respect to other factors, it is most important in a case–control study that the control group is disease-free.25 This approach has been used successfully in other studies,22,26 and was helpful during this discovery (stage 1) genomewide association study to identify the risk variants for podoconiosis. The possibility that variants involved in survival against other diseases may have been overrepresented in the controls is small, because the findings of the family-based association testing corroborated those of the genomewide association study and because HLA has no known effect on longevity.27

Identification of additional genetic risk factors reaching genomewide significance may have been limited by the small sample and the lower information content of the markers included in the SNP array used for our study population of African ancestry, as compared with populations of European ancestry.28,29 Replication of our findings in a larger sample of patients with podoconiosis and controls with the use of population-specific, denser SNP arrays is needed to inform future research directions.

Supplementary Material



Supported in part by grants from the Wellcome Trust (079791) and the Association of Physicians of Great Britain and Ireland's Links with Developing Country Scheme, as well as the Intramural Research Program of the Center for Research on Genomics and Global Health (CRGGH). The CRGGH is supported by a grant from the National Human Genome Research Institute, the National Institute of Diabetes and Digestive and Kidney Diseases, the Center for Information Technology, and the Office of the Director at the National Institutes of Health (Z01HG200362).

We thank the staff of the Mossy Foot Treatment and Prevention Association in southern Ethiopia, particularly Mr. Meskele Ashine, Mr. Zewdie Zeleke, the field nurses, and the field workers for organizing the data collection and assisting with sample collection; Prof. Heather Cordell for her thoughtful advice; and all the participants in the study.


Disclosure forms provided by the authors are available with the full text of this article at


1. Price EW. Podoconiosis: non-filarial elephantiasis. Oxford Medical; Oxford, United Kingdom: 1990.
2. Destas K, Ashine M, Davey G. Prevalence of podoconiosis (endemic non-filarial elephantiasis) in Wolaitta, Southern Ethiopia. Trop Doct. 2003;33:217–20. [PubMed]
3. Lapolla W, Tyring SK. Podoconiosis. N Engl J Med. 2011;364(12):e23. [PubMed]
4. Tekola F, Mariam DH, Davey G. Economic costs of endemic non-filarial elephantiasis in Wolaita Zone, Ethiopia. Trop Med Int Health. 2006;11:1136–44. [PubMed]
5. Yakob B, Deribe K, Davey G. High levels of misconceptions and stigma in a community highly endemic for podoconiosis in southern Ethiopia. Trans R Soc Trop Med Hyg. 2008;102:439–44. [PubMed]
6. Price EW. The association of endemic elephantiasis of the lower legs in East Africa with soil derived from volcanic rocks. Trans R Soc Trop Med Hyg. 1976;70:288–95. [PubMed]
7. Price EW, Henderson WJ. The elemental content of lymphatic tissues of barefooted people in Ethiopia, with reference to endemic elephantiasis of the lower legs. Trans R Soc Trop Med Hyg. 1978;72:132–6. [PubMed]
8. Price EW, McHardy WJ, Pooley FD. Endemic elephantiasis of the lower legs as a health hazard of barefooted agriculturalists in Cameroon, West Africa. Ann Occup Hyg. 1981;24:1–8. [PubMed]
9. Price EW. A possible genetic factor in non-filarial elephantiasis of the lower legs. Ethiop Med J. 1972;10:87–93. [PubMed]
10. Davey G, Gebrehanna E, Adeyemo A, Rotimi C, Newport M, Desta K. Podoconiosis: a tropical model for gene-environment interactions? Trans R Soc Trop Med Hyg. 2007;101:91–6. [PubMed]
11. Tekola F, Bull S, Farsides B, et al. Impact of social stigma on the process of obtaining informed consent for genetic research on podoconiosis: a qualitative study. BMC Med Ethics. 2009;10:13. [PMC free article] [PubMed]
12. Tekola F, Bull SJ, Farsides B, et al. Tailoring consent to context: designing an appropriate consent process for a biomedical study in a low income setting. PLoS Negl Trop Dis. 2009;3(7):e482. [PMC free article] [PubMed]
13. Woldegabriel G, Aronson JL, Walter RC. Geology, geochronology, and rift basin development in the central sector of the Main Ethiopian Rift. Geol Soc Am Bull. 1990;102:439–58.
14. Cesbron-Gautier A, Simon P, Achard L, Cury S, Follea G, Bignon JD. Luminex technology for HLA typing by PCR-SSO and identification of HLA antibody specificities. Ann Biol Clin (Paris) 2004;62:93–8. (In French.) [PubMed]
15. Devlin B, Roeder K. Genomic control for association studies. Biometrics. 1999;55:997–1004. [PubMed]
16. Vyse TJ, Todd JA. Genetic analysis of autoimmune diseases. Cell. 1996;85:311–8. [PubMed]
17. Mossman BT, Churg A. Mechanisms in the pathogenesis of asbestosis and silicosis. Am J Respir Crit Care Med. 1998;157:1666–80. [PubMed]
18. Ueki A, Isozaki Y, Kusaka M. Anti-caspase-8 autoantibody response in silicosis patients is associated with HLA-DRB1, DQB1 and DPB1 alleles. J Occup Health. 2005;47:61–7. [PubMed]
19. Richeldi L, Sorrentino R, Saltini C. HLA-DPB1 glutamate 69: a genetic marker of beryllium disease. Science. 1993;262:242–4. [PubMed]
20. Saltini C, Richeldi L, Losi M, et al. Major histocompatibility locus genetic markers of beryllium sensitization and disease. Eur Respir J. 2001;18:677–84. [PubMed]
Newman LS. To Be2+ or not to Be2+: immunogenetics and occupational exposure. Science. 1993;262:197–8. [PubMed]
22. de Bakker PI, McVean G, Sabeti PC, et al. A high-resolution HLA and SNP haplo-type map for disease association studies in the extended human MHC. Nat Genet. 2006;38:1166–72. [PMC free article] [PubMed]
23. Campbell MC, Tishkoff SA. African genetic diversity: implications for human demographic history, modern human origins, and complex disease mapping. Annu Rev Genomics Hum Genet. 2008;9:403–33. [PMC free article] [PubMed]
24. Tishkoff SA, Williams SM. Genetic analysis of African populations: human evolution and complex disease. Nat Rev Genet. 2002;3:611–21. [PubMed]
25. Pearson TA, Manolio TA. How to interpret a genome-wide association study. JAMA. 2008;299:1335–44. [Erratum, JAMA 2008;299:2150.] [PubMed]
26. Magnusson KP, Duan S, Sigurdsson H, et al. CFH Y402H confers similar risk of soft drusen and both forms of advanced AMD. PLoS Med. 2006;3(1):e5. [PMC free article] [PubMed]
27. Caruso C, Candore G, Colonna Romano G, et al. HLA, aging, and longevity: a critical reappraisal. Hum Immunol. 2000;61:942–9. [PubMed]
28. Spencer CC, Su Z, Donnelly P, Marchini J. Designing genome-wide association studies: sample size, power, imputation, and the choice of genotyping chip. PLoS Genet. 2009;5(5):e1000477. [PMC free article] [PubMed]
29. Teo YY, Inouye M, Small KS, et al. A genotype calling algorithm for the Illumina BeadArray platform. Bioinformatics. 2007;23:2741–6. [PMC free article] [PubMed]