Key words:exact test; Hardy-Weinberg proportions; HLA;individual test; population geneticsAcknowledgments:This work was supported in part by NationalInstitutes of Health grant GM 35326 (J.C., J.H.,S.M., G.T.), Novo Nordisk Fonden (K.S.R.),National Institutes of Health grant AI-29042(A.B.), National Institutes of Health grant GMHD25792 (M.C.K., J.J.J.), National Institutes ofHealth grant (S.M., E.H.), the National CancerInstitute of the National Institutes of Health,under contract number NO1-CO-56000 (M.C.).The content of this publication does notnecessarily reflect the views or policies of theDepartment of Health and Human Services, nordoes mention of trade names, commercialproducts, or organizations imply endorsementby the U.S. Government.Received 18 June, revised,accepted for publication 27 September 1999Copyright c Munksgaard 1999Tissue Antigens . ISSN 0001-2815Tissue Antigens 1999: 54: 533–542Printed in Denmark . All rights reserved533J.J. ChenHardy-Weinberg testing for HLA class IIJ.A. Hollenbach(DRB1, DQA1, DQB1, AND DPB1) loci inE.A. TrachtenbergJ.J. Just26 human ethnic groupsM. CarringtonK.S. RønningenA. BegovichM.-C. KingS. McWeeneyS.J. MackH.A. ErlichG. ThomsonAbstract: Testing the fit of population data to Hardy-Weinberg proportionsis crucial in the validation of many current approaches in population geneticstudies. In this paper, we tested fit to Hardy-Weinberg proportions usingexact approaches for both the overall and individual heterozygote genotypedata of four HLA Class II loci: DRB1, DQA1, DQB1, and DPB1, from 26human populations. Eighty of 99 overall tests fit the Hardy-Weinberg expec-tation (73% for DRB1, 89% for DQA1, 81% for DQB1 and 81% for DPB1).Deviations from Hardy-Weinberg proportions were both locus and group spe-cific. Although we could not rule out other mechanisms at work, the indi-vidual test results indicated that the departure was possibly partly due torecent admixture. Evidence for selection and other sources of deviation arealso discussed.The genes of the human leukocyte antigen (HLA) region controlmany important functions involved in the immune response, andare known to influence susceptibility to over 50 diseases. A numberof features of the highly polymorphic class II DR, DQ, and DPregions implicate selection as an important factor maintaining itsvariation (e.g. see (1) for review). The specific mechanism of theselection has not yet been identified, although models invoking het-erozygote advantage (since individuals heterozygous for HLA pres-ent a wider range of foreign peptides to T cells) are strong candi-dates (1–4).Many disease and population genetics studies assume Hardy-Weinberg (HW) equilibrium, or incorporate Hardy-Weinberg pro-portions (HWP) in the mathematical models. The HW law statesthat in a large random-mating population with no selection, muta-tion, or migration, allele frequencies predict genotypic frequenciesat a locus, and both genotype and allele frequencies will remainconstant in subsequent generations. Violating these assumptionsmay not always result in significant deviation from HWP. However,deviation from HWP will often imply violation of the above con-Authors’ affiliations:J.J. Chen1,J.A. Hollenbach2,E.A. Trachtenberg3,J.J. Just4,M. Carrington5,K.S. Rønningen6,A. Begovich7,M.-C. King8,S. McWeeney2,S.J. Mack3,H.A. Erlich7,G. Thomson21School of Public Health,Saint Louis University,Saint Louis, Missouri, USA,2Department of IntegrativeBiology, University ofCalifornia at Berkeley,Berkeley, California, USA,3Children’s Hospital OaklandResearch Institute, Oakland,California, USA,4Department of HumanGenetics, MilleniumPharmaceuticals,Cambridge, Massachusetts,USA,5National Cancer Institute-Frederick Cancer Researchand Development Center,Frederick, Maryland, USA,6Section of Epidemiology,National Institute of PublicHealth, Oslo, Norway,7Department of HumanGenetics, Roche MolecularSystems, Alameda,California, USA,8Departments of Geneticsand Medicine, University ofWashington, Seattle,Washington, USACorrespondence to:John J. ChenSchool of Public HealthSaint Louis University3663 Lindell Blvd.St. Louis, MO 63108USATel: π1 314 977 8134Fax: π1 314 977 8150e-mail: chenjj/slu.eduChen et al : Hardy-Weinberg testing of HLA class II dataditions and deserves further study. Checking HWP is important invalidation of many current approaches in population studies. Thisis relevant not only to theoretical population genetic or evolutionaryresearch, but also to many forensic and clinical studies. For ex-ample, a synergistic effect for the DR3 and DR4 haplotypesDRB1*0301-DQA1*0501-DQB1*0201 and DRB1*0401-DQA1*0301-DQB1*0302 which are strongly associated with type 1 diabetes inCaucasians (5), is demonstrated by deviation from HWP with excessof heterozygous versus homozygous patients.There are two important issues with regard to HWP testing. Firstis testing the overall fit of a sample. Traditionally, a large samplegoodness-of-fit c2is performed. But, due to the highly polymorphicnature of these HLA class II genes, it is not uncommon for researchersto distinguish up to 20 or 30 alleles in a sample. As a result, in manyTable 1Populations studied, their locations, sample sizes and referencesPopulation Geographic location Sample size ReferenceMixtec Alta (Amerindian) Oaxaca; Mexico 103 6Mixe (Amerindian) Oaxaca, Mexico 55 6Zapotec (Amerindian) Oaxaca, Mexico 90 6Norwegian Norway 180 7Ethiopian Ethiopia 40 8Japanese Japan 290 8Venezuelan Venezuela 27 8Mexican American USA 55 9African-American New York, New York, USA 241 10Cayapa (Amerindian) Ecuador 99 11Cauca (Colombian African American) Pacific, Colombia 20 12Choco (Colombian African American) North Pacific Coast,Colombia 20 12Providencia (Colombian African American) Caribbean Island, Colombia 30 12Coreguaje (Amerindian) Amazonas, Colombia 30 13Embera (Amerindian) Pacific, Colombia 20 13Ijka (Amerindian) Sierra Nevada, Colombia 30 13Kogui (Amerindian) Sierra Nevada, Colombia 31 13Nukak (Amerindian) Eastern Amazonas, Colombia 20 13Sikuani (Amerindian) Eastern Plains, Colombia 27 13Tule (Amerindian) NW Pacific Coast, Colombia 29 13Waunana (Amerindian) Pacific, Colombia 30 13CEPH France/Utah, USA 193 14Maya (Amerindian) Yucatan, Mexico 15 15Ticuna (Amerindian) Brazil 49 15Yanomamo (Amerindian) Venezuela 55 15Pima (Amerindian) Arizona, USA 17 16534Tissue Antigens 1999: 54: 533–542studies the genotype counts within the sample are quite small. There-fore, the large sample goodness-of-fit c2are not