High-resolution HLA allele and haplotype frequencies in several unrelated populations determined by next generation sequencing: 17^th International HLA and Immunogenetics Workshop joint report

Abstract

The primary goal of the unrelated population HLA diversity (UPHD) component of the 17^th International HLA and Immunogenetics Workshop was to characterize HLA alleles at maximum allelic-resolution in worldwide populations and re-evaluate patterns of HLA diversity across populations. The UPHD project included HLA genotype and sequence data, generated by various next-generation sequencing methods, from 4,240 individuals collated from 12 different countries. Population data included well-defined large datasets from the USA and smaller samples from Europe, Australia, and Western Asia. Allele and haplotype frequencies varied across populations from distant geographical regions. HLA genetic diversity estimated at 2- and 4-field allelic resolution revealed that diversity at the majority of loci, particularly for European-descent populations, was lower at the 2-field resolution.

Several common alleles with identical protein sequences differing only by intronic substitutions were found in distinct haplotypes, revealing a more detailed characterization of linkage between variants within the HLA region. The examination of coding and non-coding nucleotide variation revealed many examples in which almost complete biunivocal relations between common alleles at different loci were observed resulting in higher linkage disequilibrium. Our reference data of HLA profiles characterized at maximum resolution from many populations is useful for anthropological studies, unrelated donor searches, transplantation, and disease association studies.

1. Introduction

The human leukocyte antigen (HLA) genes located at 6p21.3 are amongst the most polymorphic genes in the human genome. To date, over 28,300 HLA alleles and sequences have been deposited in the IPD-IMGT/HLA Database release 3.42.0, and the number of new alleles identified is constantly increasing in an unpredictable manner [1]. This high level of allelic polymorphism, as well as the heterozygosity of HLA molecules, allows the immune system to combat the broad array of pathogens individuals may encounter [2]. On the other hand, the extreme allelic diversity can be detrimental to the success of solid-organ and hematopoietic stem cell transplantation (HSCT) between unrelated individuals; HLA allelic disparities between recipient and donor pairs increases the risk of graft rejection and graft-versus-host disease [3–5]. In addition, HLA genetic variants have been shown to associate with both susceptibility and protection to the development of many human diseases and syndromes, including various autoimmune disorders [6–8],drug-induced hypersensitivities [9,10], and cancer [11]. HLA alleles and zygosity also associate with progression and outcomes of infectious diseases [12,13]

HLA genes typically exhibit distinct allele frequencies between the major population groups originating from different geographical regions of the world [14–19]. It is now generally thought that HLA diversity between populations is maintained by balancing selection [20]. Although, it is speculated that pathogen-driven selection also may have contributed significantly to HLA variation [21]. HLA heterozygosity allows the individual to combat pathogens but diversity influences the ability of the population to combat pathogens. The signature patterns of global HLA allelic variation can be used to infer ancestry of an individual as well as to estimate ancestry proportions in admixed groups [22]. In addition, HLA allele and haplotype distributions together with linkage disequilibrium (LD) between HLA genes, which also show distinct patterns of variation between regions and diverse populations, have been exploited to examine human evolutionary processes such as migration (gene flow) and natural selection [20,23].

The study of unrelated subjects and genetic diversity by next-generation sequencing (NGS) HLA project (NGS of full-length HLA genes) of the 17^th International HLA and Immunogenetics Workshop (IHIW) was established to examine worldwide population diversity of HLA characterized at the full-gene level. Anthropological studies based on the global variation of HLA alleles have been a common theme in workshops. However, it wasn’t until the Tenth International Histocompatibility Workshop and Conference (IHWC) in 1987 [24,25], which focused on the molecular genetic basis for HLA polymorphism, that HLA population studies became a key fixture of future workshops [23,26,27]. The HLA international workshops provide an ideal opportunity to collate a vast amount of HLA data from unique populations to study human genetic variability. Although, the rapid growth of international migration in recent years has ensured that many diverse populations can be found in just a few countries [28].

During the 11–16^th workshops, HLA and other immunogenetic markers were defined using molecular technologies available at that time, such as PCR sequence-specific primer (PCR-SSP) [29], PCR reverse sequence-specific oligonucleotide probe (PCR-rSSOP) [30], PCR Sanger sequencing based typing (PCR-SSBT) [31–33]. These methods typically had limited sequence coverage because clinical laboratories only focused on the interrogation of the most polymorphic exons of the HLA gene that constituted the antigen-recognition site. Since then genomic typing methods have evolved tremendously, giving rise to the now popular next-generation sequencing (NGS) technology.

NGS was an important innovation for the molecular characterization of HLA genes since it allowed the examination of long clonal sequencing reads [34]. NGS approaches that either sequence multiple over-lapping reads encompassing partial gene segments [6,18,35–37], or sequencing multiple long reads that include whole gene amplicons [38,39] provides phase information and diminishes genotype ambiguities. In addition, the other benefits of NGS such as its amenability to unbiased variant discovery, increased accuracy, and efficiency as well as decreased costs, make NGS an attractive option for HLA typing in both the clinical and research settings. Moreover, NGS applied to population studies allows the exploration of HLA genomic population diversity on a more detailed level and provides a more accurate picture of inter- and intra-population differentiation, population admixture, and demographic history.

Here we present the findings of the analyses of HLA variation in 4,240 individuals from twelve unrelated population samples submitted to the unrelated diversity component of the 17^th IHIW. We find important differences in the distribution of allele frequencies and LD among different populations. We describe results on detailed haplotype associations observed at the maximum allelic resolution and LD measurements for 2- and 3-locus haplotypes.

2. Materials and Methods

2.1. Participating laboratories

HLA tissue typing laboratories were invited to participate by completing the ‘Study of Unrelated subjects by NGS HLA’ questionnaire available on the http://17ihiw.org/ website. The questionnaire gathered general information about the NGS protocol and instrumentation, HLA genes the laboratory could type, software packages used to analyze HLA sequencing data, and the size of the population sample(s) they planned to submit to the project. The ‘Study of Unrelated subjects by NGS HLA’ component includes large HLA genotype datasets submitted by unrelated donor bone marrow registries from Argentina (INCUCAI, >36,000 donor genotypes submitted) and Germany (DKMS, > 2 million donor genotypes submitted) as well as smaller datasets from non-registry institutions. The data from registries submitted to this component resulted from partial gene coverage in which introns were not tested. In the present study, we only included samples tested for complete class I sequences and complete or extended coverage class II loci. This study is designated the Unrelated Population HLA diversity project (UPHD), where only the non-registry data is considered. For the UPHD study, recruitment began in late 2015 and by August 2017 a total of 13 accredited histocompatibility testing laboratories from 10 different countries contributed HLA genotype and sequence data generated by various NGS protocols and instrumentations to the project. Table 1 lists the participating laboratories, principal investigators and the number of population samples submitted to the UPHD project.

Table 1.

List of participants who submitted data to the 17th IHIW Unrelated Population HLA diversity (UPHD) project.

Labcode	PI(s) name	Country	No. of Populations
areelg	Gehad EIGhazali	United Arab Emirates	1
	Zain Al Yafei
ausdes	Dianne De Santis	Australia	3
gbrcn	Colin J Brown	United Kingdom	3
gbrpep	Jennifer Pepperall	United Kingdom	1
grcsta	Catherine Stavropoulos-Giokasi	Greece	1
	Amalia Dinou
indkan	Uma Kanga	India	1
indnar	Saranya Narayan	India	1
	Srinivasan Periathiruvadi
itasac	Nicoletta Sacchi	Italy	1
	Michela Mazzocco
mexgor	Clara Gorodezky	Mexico	1
ucamor	Gerald P. Morris	United States of America	4
uilmar	Susana R. Marino	United States of America	2
umdtho	Rasmi Thomas	United States of America	4
ussta1	Marcelo A. Fernandez-Vina	United States of America	3

2.2. Population samples

The total number of population samples contributed to the UPHD project and the HLA loci typed is shown in Table 2. Initially, a total of 30 population samples typed at various loci were contributed by 13 laboratories, we excluded 1 population sample (Maori n = 7) because the sample size was less than 20 individuals. The remaining 29 population datasets were assigned to 12 worldwide ethnic groups, which are diagrammatically displayed in Figure 1; European American, African American, USA Hispanic, Mexican, Spanish, Italian, Greek, European, Arab, Asians- Pacific Islanders (API), Thai, and Indian. The final workshop dataset used for the UPHD study consists of 4,240 healthy individuals. It is important to highlight that not all populations represent the aboriginal origins of the associated geographical regions. For instance, the population groups from the USA represent self-reported ethnicity. The analyses of HLA genotype data with the double-blinded sample IDs were conducted at the Stanford Blood Center and Stanford University under the Stanford University Institutional Review Board (IRB) eProtocol titled, “17th International HLA and Immunogenetics Workshop” (#: 38899).

Table 2.

17th IHIW Unrelated Population HIA diversity (UPHD) population samples and HLA loci typed (4-Field).

	Number (2n)
Population, n	HLA-A	HLA-C	HLA-B	HLA-DRB1	HLA-DRB3/4/5	HLA-DQA1	HLA-DQB1	HLA-DPA1	HLA-DPB1
European American, 2423	4846	4796	4838	4790	4836	4712	4704	4752	4826
African American, 394	788	788	788	784	750	758	786	760	788
USA Hispanic, 56	110	112	112	112	100	100	112	100	112
Mexican, 104	208	208	208	208	208	208	208	208	208
Spanish, 276	552	552	552	544	546	544	538	550	548
Italian, 292	584	584	584	584	NT	NT	584	NT	518
Greek, 184	358	366	356	368	NT	272	366	NT	268
Arab, 52	104	104	104	104	104	104	104	104	104
Europeana, 212	424	424	424	424	394	424	424	424	424
Albanian. 2	4	4	4	4	4	4	4	4	4
Bulgarian, 1	2	2	2	2	2	2	2	2	2
Croatian. 2	4	4	4	4	4	4	4	4	4
Czech Republic, 1	2	2	2	2	2	2	2	2	2
English, 103	206	206	206	206	206	206	206	206	206
French, 1	2	2	2	2	2	2	2	2	2
German. 1	2	2	2	2	2	2	2	2	2
Irish, 1	2	2	2	2	2	2	2	2	2
Italian, 1	2	2	2	2	2	2	2	2	2
Mixed of NE ancestryb 82	164	164	164	164	164	164	164	164	164
Polish, 6	12	12	12	12	12	12	12	12	12
Portuguese, 2	4	4	4	4	4	4	4	4	4
Romanian. 9	18	18	18	18	18	18	18	18	18
APIc, 43	86	86	86	86	82	82	84	82	86
Filipino, 3	6	6	6	6	6	6	6	6	6
Japanese. 1	2	2	2	2	2	2	2	2	2
Mixed of API ancestryd 13	26	26	26	26	26	26	26	26	26
Polynesian. 1	2	2	2	2	2	2	2	2	2
Taiwanese Hakka. 1	2	2	2	2	2	2	2	2	2
USA Asian, 24	48	48	48	48	44	44	46	44	48
Thai, 42	84	84	84	84	NT	NT	84	NT	84
Indian, 162	322	324	324	324	NT	94	276	94	96

Abbreviations: 2n, total chromosome count: NT, HLA gene not typed: NE, Northern European: API, Asian Pacific Islander.

^aThe European population is comprised of the twelve populations from Europe shown in italicized text.

^bRefers to individuals mixed of entirely NE ancestry or mixed of predominantly NE ancestry.

^cThe API population is comprised of the six populations of Asian ancestry shown in italicized text.

^dRefers to individuals mixed of entirely API ancestry or mixed of predominantly API ancestry.

Figure 1.

Global location of population samples submitted to the 17^th IHIW Unrelated Population HLA Diversity (UPHD) project.

2.3. Submission of HLA data into the 17th IHIW database

The 17^th IHIW database https://ihiws17.stanford.edu was the central storage point for all workshop data [40]. In the 17th IHIW database, participating laboratories completed a more detailed picture about each sample submitted to the project, including geographical, demographic and linguistic information, as well as detailed parameters of the NGS reagents, instrumentation and software used. Laboratories submitted HLA data by uploading Histoimmunogenetics Markup Language (HML) or eXtensible Markup Language (XML) typing report files to the 17^th IHIW database. The IPD-IMGT/HLA Database version 3.25.0 (released July 2016) [1] was used as the only reference source for the 17^th IHIW. If the alleles submitted by laboratories were analyzed using other reference database versions, the workshop database converted those allele names to correspond to their most similar lowest digit allele present in version 3.25.0. All HLA genotype data was converted to the genotype list (GL) string format [41,42] in the workshop database in order to standardize input data and facilitate downstream analyses.

2.4. HLA typing

All samples included in the UPHD project were typed for HLA loci using the established NGS methods used routinely by the participating laboratories. Two core laboratories (labcodes ussta1 and utxask) were identified to perform NGS HLA typing on samples submitted by three laboratories; labcodes mexgor, areelg, and gbrpep. For the remaining laboratories, all operations from PCR amplification through to generation of HLA data were performed at the individual participant sites. Various NGS methods were employed, but most of the participating laboratories shared common NGS reagents and protocols, NGS software, and hardware. All groups used NGS commercial reagents and protocols for typing: NGS Engine (GenDx, Utrecht, Netherlands); TypeStream (One Lambda/Thermo Fisher Scientific Inc., CA, USA); Holotype HLA & HLA Twin (Omixon, Budapest, Hungary); MIA FORA NGS (Immucor, Inc., Norcross, GA, USA); TruSight HLA Assign (Illumina Inc. CA, USA). The number of HLA loci typed at the full-gene level for class I loci and full-gene or wide coverage for class II genes ranged from 5 to 11 across the groups. A summary of the NGS software and hardware, as well as HLA loci typed by each group is shown in Table 3.

Table 3.

NGS systems and HLA Loci typed by participating laboratories.

Labcode	Software	Hardware	HLA loci typed	Coverage class I/class II
areelg*	NGSengine, GenDx	MiSeq, Illumina	A, C, B, DRB1/3/4/5, DQA1, DQB1, DPA1, DPBI	Full-gene/wide-coverage
ausdes	NGSengine, GenDx	Ion S5™ system	A, C, B, DRB1/3/4/5, DQB1, DPB1	Full-gene/wide-coverage
gbrcn	TypeStream™ Visual, One lambda Thermo Fisher Scientific	Ion S5™ system	A, C, B, DRB1/3/4/5, DQA1, DQB1, DPA1, DPBI	Full-gene/full-gene (DQA1, DPA1) and wide-coverage
gbrpep*	NGSengine, GenDx	MiSeq, Illumina	A, C, B, DRB1/3/4/5, DQA1, DQB1, DPA1, DPBI	Full-gene/wide-coverage
grcsta	HLA Twin, Omixon	MiSeq, Illumina	A, C, B, DRB1, DQA1, DQB1, DPB1	Full-gene/wide-coverage
indkan	NGSengine, GenDx	MiSeq, Illumina	A, C, B, DRB1, DQA1, DQB1, DPA1, DPB1	Full-gene/wide-coverage
indnar	MIA FORA FLEX, Immucor	MiniSeq, Illumina	A, C, B, DRB1, DQB1	Full-gene/wide-coverage
itasac	NGSengine, GenDx	MiSeq, Illumina	A, C, B, DRB1, DQB1	Full-gene/wide-coverage
mexgor‡	MIA FORA FLEX, Immucor	MiSeq, Illumina	A, C, B, DRB1/3/4/5, DQA1, DQB1, DPA1, DPB1	Full-gene/wide-coverage
ucamor	TruSight HLA Assign, Illumina	MiSeq, Illumina	A, C, B, DRB1/3/4/5, DQA1, DQB1, DPA1, DPBI	Full-gene (A, C), B (exon1-intron 6) /full-gene (DQ DPA1) and wide-coverage
uilmar	NGSengine, GenDx	MiSeq, Illumina	A, C, B, DRB1/3/4/5, DQA1, DQB1, DPA1, DPBI	Full-gene/wide-coverage
umdtho	NGSengine, GenDx	MiSeq, Illumina	A, C, B, DRB1, DQB1, DPB1	Full-gene/wide-coverage
ussta1	MIA FORA FLEX, Immucor	MiniSeq and NextSeq, Illumina	A, C, B, DRB1/3/4/5, DQA1, DQB1, DPA1, DPB1	Full-gene/wide-coverage

^*Samples typed by laboratory utxask (Baylor University Medical center).

^‡Samples typed by laboratory ussta1 .

2.5. Data pre-processing and checking

2.5.1. HLA allele ambiguity assignments

Population allele data were systematically reviewed and pre-processed or ‘cleaned’ to assign common ambiguities found across groups. The majority of the ambiguities were observed in class II alleles which were due to the presence of short tandem repeat (STR) enriched regions located within intronic gene segments of some genes. These low complexity regions, consist of ~1–6 bp nucleotide units that are repeated numerous times and cannot be enumerated with high accuracy by the NGS method. In an attempt to standardize alleles that could not be discriminated due to STRs, alleles were assigned to groups and were given the suffix SG (STR Group) to the lowest numbered allele in that group [6,37]. For instance, HLA-DRB1*15:01:01:01SG denotes the HLA-DRB1*15:01:01:01/HLA-DRB1*15:01:01:02/HLA-DRB1*15:01:01:03 STR ambiguous group. Characteristics of ambiguities due to STRs and non-sequenced regions identified in the UPHD dataset are shown in Supplementary Table 1.

2.5.2. HLA-DRB1~HLA-DRB3/4/5 haplotypes

The occurrence of functional HLA-DRB (HLA-DRB3, HLA-DRB4, HLA-DRB5) genes on haplotypes varies according to the presence of specific HLA-DRB1 alleles [43]. HLA-DRB3 occurs whenever the HLA-DRB1*03, 11, 12, 13, 14 alleles are present. Similarly, HLA-DRB4 is present with HLA-DRB1*04, 07, 09 alleles, whilst HLA-DRB5 is found on haplotypes bearing HLA-DRB1*15, 16 alleles. Typically, if the HLA-DRB1*01, 08, 10 alleles are present all of the functional HLA-DRB genes are absent. Each individual may contain zero, one, or two copies of a HLA-DRB3, HLA-DRB4, and HLA-DRB5 allele. The absence of HLA-DRB3/4/5 genes on HLA-DRB1*01, 08, 10 bearing haplotypes were assigned HLA-DRB*00:00 and were counted when calculating allele frequencies; if blank alleles are not considered the frequencies of alleles at the HLA-DRB3/4/5 loci will be over-estimated and deviations from HWE may be observed under the assumption that the presence of a single allele corresponds to a homozygous genotype at this combined locus. HLA-DRB haplotypes have been well-defined across many populations, however, there are reported exceptions such as HLA-DRB1*08~HLA-DRB3 found in Asian populations [44] and HLA-DRB1*15~HLA-DRB5*absent haplotypes in African descent populations [18]. In the present study HLA-DRB data that deviated from the general pattern of HLA-DRB3/4/5~HLA-DRB1 associations mostly resulted from allele drop-out or non-typing of a particular HLA-DRB gene. For instance, a group may consistently type HLA-DRB3 and HLA-DRB4 but not HLA-DRB5 in this case genotype data were excluded from any subsequent haplotype analyses involving the HLA-DRB3/4/5 genes. In this study, the HLA-DRB3, HLA-DRB4, HLA-DRB5 loci were evaluated as a single combined locus, denoted HLA-DRB3/4/5, in order to increase the accuracy of the low frequency haplotypes estimated by the EM programs.

2.6. Population statistical analyses

The Python for Population Genomics (PyPop) v0.7.0 software package [45] was used to perform the majority of the population analyses. Allele carrier frequencies were determined by direct counting and were calculated as the number of individuals carrying a specific allele (either at the homozygous or heterozygous state) divided by the total number of individuals. Allele frequencies at each locus were tested for deviations from Hardy-Weinberg equilibrium (HWE) proportions using the exact test of Guo and Thompson [46] and by the chi-square test when expected counts were equal or greater than 5.

An implementation of the iterative Expectation-Maximization (EM) algorithm in the PyPop software was used to estimate the frequencies of two loci allelic haplotypes. Linkage disequilibrium (LD) between pairs of alleles at different loci and overall ‘global’ LD between pairs of loci were calculated. Overall LD was computed using two formulae’s normalized to range from −1 to +1; Hendricks D_ij’ statistic [47] which weights the contribution to LD of specific allele pairs by the product of their allele frequencies, and Cramér’s V statistic [48] also described as W_n that calculates a chi-squared statistic for deviations between observed and expected haplotype frequencies, and conditional asymmetric LD (cALD) measures. cALD between two HLA loci describes the level of allele variation at locus 1 given the presence of specific alleles at locus 2. For example, it is possible that a specific allele at locus 1 is in complete LD with a specific allele at locus 2, such as HLA-DQA1*05:01:01:02 (at locus 1) associates exclusively with HLA-DQB1*02:01:01 (at locus 2). However, HLA-DQB1*02:01:01 also associates with HLA-DQA1*05:01:01:01 and HLA-DQA1*05:01:01:03 therefore LD for both of these haplotypes is less than 1. This complimentary pair of cALD measures is denoted W_loc1/loc2 and W_loc2/loc1, when both these measures are equal to 1, meaning a complete correlation between both loci, the ALD measure is equal to W_n. cALD was computed using the formula described by Thomson and Single [49] implemented in the Phased Or Unphased Linkage Disequilibrium (POULD) v0.9.1.9000 package.

Extended haplotype (encompassing more than two loci) frequencies were estimated using the R ‘haplo.stats’ package to run EM in the Bridging Immuno Genomic Data Analysis Workflow Gaps (BIGDAWG) v1.8 package [50]. To reduce the uncertainty of extended haplotypes estimated, particularly those encompassing more than six loci, we limited our analyses to individuals that had complete allelic genotype data.

3. Results

For each ethnic group submitted to the UPHD project, allele frequencies estimated at the maximum allelic resolution, LD measurements between pairs of loci, and 2-, 3-, 4-, 5-, 9-locus haplotype frequencies are available on the http://17ihiw.org website.

3.1. Hardy-Weinberg equilibrium (HWE) tests

The results of HWE exact tests performed at each locus in the 12 population samples are summarized in Table 4. The maximum number of loci tested was 3 class I and 6 class II loci, since HLA-DRB3, HLA-DRB4, and HLA-DRB5, were evaluated as a combined locus denoted HLA-DRB3/4/5. All data was tested at the original maximum allelic resolution and 2-field resolution. At the maximum resolution, significant HWE deviations (P<0.05) were observed for at least one locus amongst 11 population samples. In the Greek and Thai populations, no significant differences from HWE expectations were detected at any of the genotyped loci.

Table 4.

Probability values for Guo and Thompson Hardy-Weinberg equilibrium (HWE) tests in the UPHD populations.

Maximum allelic resolutiona
Population	HLA-A	HLA-C	HLA-B	HLA-DRB1	HLA-DRB3/4/5	HLA-DQA1	HLA-DQB1	HLA-DPA1	HLA-DPB1
European American	0.0036	0.0001	0.0000	0.0001	0.0000	0.0067	0.0014	0.0000	0.0000
African American	0.3967	0.0458	0.0004	0.0002	0.0045	0.0002	0.0010	0.0457	0.1030
USA Hispanic	0.0034	0.0011	0.0977	0.1916	0.0107	0.2181	0.0879	0.0227	0.0331
Mexican	0.2713	0.0491	0.0629	0.2286	0.0806	0.0375	0.0183	0.2912	0.1346
Spanish	0.5230	0.9906	0.1151	0.8920	0.5300	0.3015	0.7955	0.0129	0.9012
Italian	0.8447	0.0985	0.8679	0.2448	NT	NT	0.1224	NT	0.0025
Greek	0.4691	0.5181	0.7447	0.3585	NT	0.1268	0.2305	NT	0.1233
Arab	0.0007	0.1866	0.1187	0.2123	0.5325	0.1640	0.1951	0.3769	0.4846
European	0.9083	0.2381	0.0852	0.0075	0.0082	0.0012	0.0068	3.0E-05	0.1332
API	0.0005	6.0E-05	0.0000	0.0003	0.0009	0.0001	0.0226	0.1848	0.2367
Thai	0.8839	0.0970	0.3526	0.2229	NT	NT	0.2942	NT	0.7104
India	0.2216	0.8316	0.2299	0.4482	NT	0.4661	0.1402	0.0219	0.0197
2-Field allele resolutionb
Population	HLA-A	HLA-C	H LA-B	HLA-DRB1	HLA-DRB3/4/5	HLA-DQA1	HLA-DQB1	HLA-DPA1	HLA-DPB1
European American	0.0024	0.0053	2.0E-05	0.0001	0.0005	0.1575	0.0027	0.0007	0.0000
African American	0.5170	0.0446	0.0004	0.0003	0.0122	0.0969	0.0491	0.5912	0.1328
USA Hispanic	0.0041	0.1008	0.0511	0.1217	0.3845	0.1598	0.2138	0.3784	0.0249
Mexican	0.3126	0.0598	0.0835	0.1564	0.2390	0.0181	0.0641	0.0713	0.5244
Spanish	0.3135	0.9281	0.2699	0.8588	0.2817	0.7791	0.5180	0.1513	0.8987
Italian	0.8016	0.2731	0.6705	0.2305	NT	NT	0.2882	NT	0.0004
Greek	0.3984	0.8091	0.8724	0.3815	NT	0.3061	0.1838	NT	0.0579
Arab	0.0003	0.2089	0.1479	0.1720	0.6508	0.4250	0.1559	0.7203	0.3533
European	0.7474	0.1366	0.1881	0.0039	0.5274	0.0340	0.0016	0.7028	0.1569
API	0.0005	0.0019	0.0000	0.0004	0.0624	0.0112	0.0430	0.5365	0.2509
Thai	0.8633	0.0970	0.3526	0.2229	NT	NT	0.2942	NT	0.6943
Indian	0.4011	0.5449	0.6403	0.3584	NT	0.5607	0.1448	1.0000	0.0197

HLA loci that deviated significantly (P < 0.05) from HWE are shown in boldface type.

^aHWE test performed on alleles characterized at maximum (3–4-Field) allelic resolution.

^bHWE test performed on alleles characterized at 2-Field allele resolution.

For the European American sample, all loci tested at maximum resolution exhibited deviations from HWE. Analyses using 2-field allelic resolution data only eliminated HWE deviation at the HLA-DQA1 locus. Most of the HWE deviations could be attributed to an excess of specific homozygous and heterozygous genotypes at 5 of these loci (HLA-A, HLA-C, HLA-DPA1, HLA-DPB1, HLA-DRB3/4/5) and mostly an excess of heterozygous genotypes at the other 4 loci (HLA-B, HLA-DQA1, HLA-DQB1, HLA-DRB1). It is noteworthy that low frequency or rare alleles were detected in the European American samples that contributed to some of the HWE deviant genotypes. Importantly, according to the allele frequencies database (ww.allelefrequencies.net) and other sources these infrequent or rare alleles have been observed in elevated frequencies in other ethnic groups such as Ashkenazi Jews (HLA-A*03:02, HLA-A*26:08), and Natives from the American continents (HLA-B*15:15, and HLA-B*39:05:01) [51]. The presence of these alleles strongly suggests subpopulation structure also known as the Wahlund effect [52], which occurs due to admixture of two or more populations with distinct allele frequencies resulting in overall decreased heterozygosity of the admixed population. It is also possible that genotyping errors may have also contributed to HWE deviations. Another factor to be considered is that the European American or ‘USA white’ is a loosely based ethnic category defined by the USA 2010 census since it encompasses individuals having origins from vast and distant geographic regions including Europe, North Africa, and the Middle East [53].

Similarly, in the African American population, extreme and moderate deviations from HWE proportions at HLA-DPA1, HLA-DQA1, and HLA-DRB3/4/5 were the results of roughly equal proportions of excess observed homozygous and heterozygous genotypes compared to the expected genotypes. For the remaining loci that deviated from HWE a significant excess of heterozygous genotypes were observed. Again, population substructure was evident due to the presence of rare alleles that are common in other populations.

An excess of specific heterozygous genotypes accounted for all of the loci that deviated from HWE proportions in the USA Hispanic (HLA-A, HLA-C, HLA-DRB3/4/5, HLA-DPA1, HLA-DPB1), Mexican (HLA-C, HLA-DQA1, HLA-DQB1), API (all loci except HLA-DPA1 and HLA-DPB1), and Indian (HLA-DPA1, HLA-DPB1) population samples. Whereas, in European’s significant deviations at HLA-DPA1, HLA-DQB1, and HLA-DRB3/4/5 were due to both excess homozygous and heterozygous genotypes, and excess heterozygous genotypes explained the deviation observed at HLA-DQA1 and HLA-DRB1.

In general, for all populations, HWE testing using 2-Field allelic data did little to correct the majority of deviations observed using maximum resolution data. For this reason, other population parameters presented are based on 3/4-Field allelic resolution data.

3.2. Genetic Diversity

The genetic diversity within each population was assessed using the expected heterozygosity index values per locus, the results of which are displayed in Figure 2. Supplementary Table 17 lists the observed and expected heterozygosity values. For all class I loci the average heterozygosity index was highest, at near maximal, for HLA-B ranging from 0.94 to 0.97 across all populations; the largest value was observed in USA Hispanics.

Figure 2.

Column plots of expected heterozygosity values for HLA class I and II loci (4-Field data).

Overall, heterozygosity was greater at HLA-C (0.88 to 0.94) than HLA-A (0.85 to 0.94). From the HLA-A heterozygosity values, we see clear differences between populations with the highest values detected in African American and USA Hispanics, intermediate values observed in the API, Spain, and Italy and the lowest values seen in Europeans, European Americans, and the remaining Asian populations. On average, the USA Hispanics exhibited the highest heterozygosity values at class I loci, followed by African Americans, and Arabs. It is noteworthy that the USA Hispanic population represents an amalgam of populations from North, Central and South America. Also, the high level of diversity observed in USA Hispanics is reflective of their multi-ethnic composition which includes contributions from European, Native American, and African populations. Population substructures within this multi-ethnic group may also explain the HWE deviations detected.

It is well known that African populations are highly genetically heterogeneous; this is likely due to both an ancient complex demographic history, and pathogen-driven selection acting in response to unusual pathogens [54,55]. Geographic separation of tribes along with common pathogens also play a role in HLA diversification on the continent. These features, as well as admixture with non-African populations, may account for the relatively high HLA diversity observed in the African American population.

At the class II loci, expected heterozygosity was highest at the HLA-DRB1 locus; the highest value of 0.95 was observed in USA Hispanics, followed by 0.94 in African Americans and Arabs. The lowest value of 0.90 was observed in the API sample. Due to the fact that only 8 groups typed the HLA-DRB3/4/5 loci, it was difficult to observe trends in heterozygosity values across populations. Interestingly at the HLA-DQA1 locus, the lowest value of 0.87 was observed in both the USA Hispanic and Mexican populations, whereas the highest value of 0.92 was found in API’s. In the USA Hispanic and Mexican populations the relatively low values are due to predominance of just a few alleles in the relatively small (n ~20) HLA-DQA1 allele repertoire such as HLA-DQA1*03:01:01, HLA-DQA1*04:01:01, and HLA-DQA1*05:051:01:01. In comparison, the API group consisted of several HLA-DQA1 alleles of intermediate frequencies, accounting for the overall high heterozygous frequencies. The HLA-DQB1 heterozygosity values across groups displayed a similar pattern to the HLA-DQA1 values. At the HLA-DPA1 gene the lowest value was seen in Mexicans (0.69), due to the high frequency of the HLA-DPA1*01:03:01:05 (allele frequency, AF = 0.52) and the highest values were seen in the African American (0.88) and the Asian Indian (0.86) groups. A similar trend was observed at the HLA-DPB1 locus.

In summary, USA Hispanics and African American populations are more genetically diverse at the majority of the HLA loci, than European origin (intermediate levels of diversity) and Asian populations (lowest level) populations.

3.3. Population differences in allele frequencies

The number of unique HLA alleles identified in each population is shown in Table 5. For all populations, the largest number of unique alleles was observed at HLA-B and the lowest at HLA-DPA1. For populations typed at 4-field resolution, there was a direct correlation with the sample size and the number of unique alleles identified, with European Americans having the largest number of unique alleles and the smallest number found in the Thai group. We observed different levels of variation across populations in regards to HLA class I and class II allele frequencies. In general, class II alleles exhibited lower levels of diversity compared to class I alleles. The African American and USA Hispanic populations had the most variation, at both class I and II loci; the least variation was observed in the Asian populations. Overall, 743 unique alleles were identified in the total 12 populations that characterized alleles at maximum resolution; HLA-A (118), HLA-C (109), HLA-B (197), HLA-DPA1 (23), HLA-DPB1 (68), HLA-DQA1 (43), HLA-DQB1 (57), HLA-DPB1 (97), HLA-DRB3/4/5 (29). Class I and class II allele frequencies estimated per population are presented in Supplementary Table 2.

Table 5.

The number of unique HLA alleles per population.

Number of unique alleles (k)
Population	HLA-A	HLA-C	HLA-B	HLA-DRB1	HLA-DRB3/4/5	HLA-DQA1	HLA-DQB1	HLA-DPA1	HLA-DPB1
European American	70	72	107	59	24	36	33	23	45
African American	50	50	71	47	14	28	27	18	41
USA Hispanic	30	30	43	31	13	20	20	11	20
Mexican	27	29	46	34	13	19	21	14	16
Spanish	36	40	53	37	14	23	24	14	28
Italian	43	41	58	38	NT	NT	24	NT	32
Greek	36	37	58	33	NT	25	22	NT	22
Arab	30	26	36	28	12	19	17	11	18
European	45	43	58	38	17	25	24	13	25
API	18	21	30	23	13	19	16	9	17
Thai	17	16	29	23	NT	NT	12	NT	19
Indian	34	32	54	33	NT	19	29	11	18

Abbreviations; NT; Loci not typed.

3.3.1. Class I allele frequencies

3.3.1.1. HLA-A

HLA-A*02:01:01:01 was the most common allele identified in the entire cohort, with an overall allele frequency (AF) of 0.233. This allele was the top ranking in 9 populations, with frequencies ranging from over 0.316 in Europeans and 0.127 in USA Hispanics. Three other 4-field variants of HLA-A*02:01:01 allele were identified, however they occurred at much lower frequencies amongst selected groups, HLA-02:01:01:03 (European American, AF = 0.0006), HLA-02:01:01:04 (European American, AF = 0.0002), and HLA-02:01:01:05 (European American, AF = 0.0008; Europeans AF = 0.0024). In comparison, in the API, Thai, and Indian populations, HLA-A*11:01:01:01 (API AF = 0.174, Thai AF = 0.286), and HLA-A*01:01:01:01 (India AF = 0.165) were the most common. Other common alleles identified in populations originating from European regions such as Northern European, South-West Europe (Spain), Southern Europe (Italy), Southern Eastern Europe (Greece), as well as European Americans were HLA-A*01:01:01:01 (AF = 0.106 – 0.165), HLA-A*03:01:01:01 (AF = 0.067 – 0.133), HLA-A*24:02:01:01 (AF = 0.040 – 0.137), and HLA-A*11:01:01:01 (AF = 0.040 – 0.085). However, the USA Hispanics and Mexican populations also shared some of the common alleles identified in the populations originating from European regions, for instance, allele HLA-A*24:02:01:01 was the second most common in these groups. In the African American population, allele HLA-A*30:01:01 was common (AF = 0.084) but was found at lower frequencies (AF = 0.012 – 0.029) in all other groups. HLA-A*74:01:01 was only detected in both the African (AF = 0.049) and Arab (AF = 0.038) populations. The Arab samples were from the indigenous United Arab Emirates (UAE). The UAE population structure is diverse due to admixture of the indigenous population with immigrants from Yemen, Oman, North Africa, India, Europe, Australia, North America, and Latin America. Another region-specific allele found at an elevated frequency (~0.05) was HLA-A*02:11:01 found only in Asian Indians (AF = 0.071).

3.3.1.2. HLA-B

Overall, the most common HLA-B allele identified was HLA-B*07:02:01, occurring at frequencies ranging from 0.041 to 0.127. However, this allele was found at a lower frequency in Arabs (AF = 0.0096) and was not detected in the Thai group. The other common alleles HLA-B*08:01:01:01, HLA-B*44:02:01:01, HLA-B*35:01:01:02, and HLA-B*15:01:01:01 were found predominantly in individuals of European ancestry, as well as the Mexican and Indian populations. HLA-B*53:01:01 was the top-ranked allele in African Americans, occurring at frequencies of 0.114, in comparison this allele was observed at either very low frequencies, <0.01, or not at all in the other groups. This finding suggests that HLA-B*53:01:01 can be used as a marker to distinguish African populations from other worldwide populations. Similarly, HLA-B*39:05:01 was found at high frequencies in Mexicans (AF = 0.111) and USA Hispanics (0.089), but was observed at very low frequencies (<0.0025) in 3 groups (European American, African American, Italian), and not detected in the remaining groups. HLA-B*39:05:01 is an allele frequently observed in individuals of Native American ancestry. Noteworthy is the allele HLA-B*39:02:02 which was found solely in the Mexican group at a relatively high frequency of 0.082 but was not detected in USA Hispanics as well as the other groups. The Mexicans in this study correspond to a relatively isolated group and may not be representative of the general Mexican population or Hispanic group [51]. The API and Thai group shared some common HLA-B alleles, however, there were distinct differences in frequencies. For example, the allele frequency of HLA-B*46:01:01 was 0.070 in the API group, but 0.143 in the Thai group. Also, HLA-B*40:06:01:01 AF = 0.105% in API, AF = 0.024 in Thai’s. In contrast HLA-B*38:02:01 is lower in APIs (AF = 0.023) but higher in Thai’s (AF = 0.071). Two alleles, HLA-B*15:12 (AF = 0.058), HLA-B*15:21 (AF = 0.035) were exclusive to the API group. In general populations with the largest number of alleles had a larger number of rare alleles.

3.3.1.3. HLA-C

Globally, HLA-C*07:01:01:01 was the most frequent allele observed at 0.121 and was detected across all groups with the highest frequency occurring in Europeans (AF = 0.149) and lowest in the Asian and Arab groups, AF = ~0.009 – 0.023. The 4-field alternative, HLA-C*07:01:01:04, was only observed in the Indian group, albeit at a very low frequency; 0.0093. Similarly, HLA-C*07:02:01:03 was also relatively common across all groups (global AF = 0.095), with the exception of the Thai group where it was not observed, and the lowest frequencies found in the African ancestry groups; African Americans (AF = 0.0203), and Arabs (AF = 0.010). In comparison, the 4-field counterpart, HLA-C*07:02:01:01 was observed at higher frequencies in African Americans (0.046), Arabs (0.125), API (0.163), Thai (0.131), and an extremely elevated frequency of 0.260 in Mexicans. Countries from North and South Europe had lower frequencies of around 0.01.

Three 4-field variants of HLA-C*04:01:01 were detected in the entire cohort. HLA-C*04:01:01:01 was the most frequent in the African American group at (AF = 0.193), followed closely by the Mexicans (AF = 0.183), and ~ 0.070 to 0.120 in the remaining groups. The highest frequency of HLA-C*04:01:01:06 was found in Italians (AF = 0.079), and approximately 0.040 in Spanish, Greeks, and Arabs. Overall, HLA-C*04:01:01:05 was less common with frequencies ranging from 0.005 to 0.015 across eight groups; this allele was not detected in USA Hispanics, APIs, Thais, and Arabs.

The HLA-C*06:02 allele group was also relatively common in the entire cohort. HLA-C*06:02:01:01 was present in all groups; AF ranging from 0.018 in USA Hispanics to 0.170 in Indians. HLA-C*06:02:01:02 was the top-ranking allele in the Arab group (AF = 0.135), but occurred at either low frequencies across the other groups or were not detected at all. Likewise, allele HLA-C*06:02:01:03 was either absent or not very common amongst the population samples.

3.3.2. Class II allele frequencies

3.3.2.1. HLA-DPA1 and HLA-DPB1

The five intronic variants of HLA-DPA1*01:03:01 were the most common, collectively, accounting for 76% of the total HLA-DPA1 allelic diversity observed across all groups. Overall, HLA-DPA1*01:03:01:02 was the most common variant across groups (AF = 0.154 – 0.307), with the exception of the following groups: USA Hispanics, where HLA-DPA1*01:03:01:05, was highly frequent (AF = 0.40); HLA-DPA1*01:03:01:01 was the most common in India (AF = 0.202) and APIs (AF = 0.293); HLA-DPA1*01:03:01:04 Arabs (AF = 0.240); HLA-DPA1*01:03:01:05 in Mexicans (AF = 0.524).

The HLA-DPA1*02:01:01:01 and HLA-DPA1*02:01:01:02 occurred at relatively high frequencies globally, 0.050 and 0.065 respectively. Strikingly, HLA-DPA1*02:01:01:01 was most frequent in African Americans, USA Hispanics, Europe, Spanish (AF = ~0.08 – 0.112), whereas HLA-DPA1*02:01:01:02 was common amongst Mexicans, APIs, and Arabs (AF = 0.05 – 0.135).

The most frequent HLA-DPB1 allele detected was the 4-field ambiguous pair HLA-DPB1*04:01:01:01/HLA-DPB1*04:01:01:02, representing over 0.32 in countries from Europe, 0.375 in Indians and 0.356 in Arabs. In African Americans HLA-DPB1*01:01:01 was most common (AF = 0.237), whereas HLA-DPB1*04:02:01:02 ranked number 1 in USA Hispanics (AF = 0.366) and Mexicans (AF = 0.510). HLA-DPB1*04:01:01:01/HLA-DPB1*04:01:01:02 and HLA-DPB1*02:01:02/HLA-DPB1*02:01:19 was equally common in the API group (AF = 0.174). Also HLA-DPB1*03:01:01 was relatively common across groups, ~0.02– 0.08. HLA-DPB1*18:01 was found at a relatively high frequency in African Americans (AF = 0.074), observed on one occasion in the USA Hispanic and Arab populations, but absent in the other groups; this allele occurs at ~ 5% in West, Central, and South African populations (www.allelefrequencies.net).

3.3.2.2. HLA-DQA1 and HLA-DQB1

For HLA-DQA1, the ambiguous intronic SG-level alleles were very common and accounted for 57.5% of the total allele frequency.

The HLA-DQA1*01:02:01:01SG allele was found in all groups, at AF around 0.04 to 0.208. Whereas the HLA-DQA1*01:02:01:04SG and HLA-DQA1*01:02:01:02 intronic counterparts were less common. Similarly, HLA-DQA1*01:03:01:02SG (AF = 0.053) was more common than HLA-DQA1*01:03:01:01 (AF = 0.013) and HLA-DQA1*01:03:01:03SG (AF = 0.0006). HLA-DQA1*02:01:01:01SG was common in European Americans (AF = 0.126), Europeans (AF = 0.170), Spanish (AF = 0.162), Indians (AF = 0.277), and Arabs (AF = 0.144). Whereas, HLA-DQA1*01:02:01:01SG was most common in African Americans (AF = 0.208), HLA-DQA1*05:05:01:01SG in Greeks (AF = 0.254) and Arabs (AF = 0.164). Whilst HLA-DQA1*03:01:01 was most frequent in USA Hispanics (AF = 0.31) and Mexicans (AF = 0.274).

Of the total 57 HLA-DQB1 alleles characterized globally HLA-DQB1*06:02:01 was the most common occurring at a frequency of 11.1%. This allele was also the most frequent in European Americans (AF = 0.123), African Americans (AF = 0.219), and Europeans (AF = 0.144).

Globally HLA-DQB1*02:01:01 was the second most frequent allele (AF = 0.1047) closely followed by HLA-DQB1*03:01:01:03 (AF = 0.1029). The latter allele was prevalent in the Southern European countries; Italy AF = 0.236 and Greece AF = 0.2477. In comparison, the 4-field alternative HLA-DQB1*03:01:01:01 was found at elevated frequencies in the API and Thai populations (0.214). Whereas, HLA-DQB1*03:01:01:02 was more prevalent in Arab’s (AF = 0.135).

HLA-DQB1*06:01:01 was most frequent in API and India, AF of 0.131 and 0.156 respectively. HLA-DQB1*03:19:01 was the second most frequent in African Americans at 0.093, and was detected at a moderate frequency in Arabs (AF = 0.039), and very low frequencies in European Americans (AF = 0.002), USA Hispanics (AF = 0.018), Spanish (AF = 0.0074), Mexicans (AF = 0.005), and Italians (AF = 0.005). HLA-DQB1*03:19:01 is a well-known HLA allele of African ancestry, suggesting African gene flow in the aforementioned populations.

3.3.2.3. HLA-DRB1, HLA-DRB3, HLA-DRB4, and HLA-DRB5

At the HLA-DRB1 locus, overall alleles HLA-DRB1*07:01:01:01SG, HLA-DRB1*03:01:01:01SG, and HLA-DRB1*15:01:01:01SG were the most common occurring at allele frequencies above 0.10. These alleles were relatively common among all groups from Europe, Thai, India, and Arab. Whereas, HLA-DRB1*15:03:01:01SG is common in African Americans at 0.149, and were found on single occasions in European Americans, Mexicans, and Arabs. HLA-DRB1*04:07:01 was common to USA Hispanics (AF = 0.116) and Mexicans (0.149), as well as HLA-DRB1*14:06:01; USA Hispanics, 0.036 and Mexican 0.063 but was found at lower frequencies in other groups. HLA-DRB1*14:05:01 was exclusive to the API group, AF = 0.047.

There are eight different alleles in the HLA-DRB1*08 family, and they all exhibit varying frequencies across the major population groups. For instance, HLA-DRB1*08:01:01 is present at ~ 0.02 in European Americans, Europeans, Spanish, and Italians. Whilst HLA-DRB1*08:04:01 was most common in African Americans (AF = 0.063) and Arabs (AF = 0.019). HLA-DRB1*08:02:01 was frequent in Mexicans (AF = 0.130) and USA Hispanics (AF = 0.063). In Asians, HLA-DRB1*08:03:02 was most moderately common; API (AF = 0.035), Thai (AF = 0.024) but was either absent or found at lower frequencies in the other groups.

At the HLA-DRB3 locus, the most frequent allele, globally, was HLA-DRB3*02:02:01:02 (0.140), the 4-field alternative form HLA-DRB3*02:02:01:01 occurred less frequently at 0.063. At HLA-DRB4, the HLA-DRB4*01:03:01:01/HLA-DRB4*01:03:01:03 ambiguous allele pair was the most frequent across all groups at ~0.113 to 0.275. The most frequent HLA-DRB5 allele detected was HLA-DRB5*01:01:01 (global AF = 0.127). Due to the limited allelic diversity at the HLA-DRB3, HLA-DRB4, HLA-DRB5 loci, there were no obvious distribution patterns observed amongst the populations.

3.4. Population differences in the distribution of Linkage disequilibrium and 2-locus haplotypes

3.4.1. Global (locus-level) linkage disequilibrium

The results of global LD tests to assess the strength of association between neighboring and non-neighboring loci pairs are summarized in Supplementary Table 3. Due to deviation from HWE, the LD and haplotype data presented in this manuscript should be used with caution. Three measures of LD were used for alleles characterized at maximum and 2-field resolution; D’, W_n, W_loc1/loc2, W_loc2/loc1. The standard D’ and W_n values range from 0 to 1, 0 refers to linkage equilibrium, and positive values indicate association; values >0.8 suggests a very strong association. There is a strong negative relationship between the measure of LD, as well as the recombination fraction, and the physical distance separating the loci on the chromosome; generally longer inter-marker distances are associated with smaller LD values. LD plots displaying D’ and W_n measurements for class I and class II haplotypes are depicted in Figures 4 and 5 respectively.

Figure 4.

Global linkage disequilibrium values of D’ and W_n class II haplotypes by population.

Overall, strong associations (D’ values greater than 0.5) were observed for all loci pairs except for non-neighboring loci pairs that included either the HLA-A or HLA-DPB1 loci that showed weak LD. However, unexpectedly the USA Hispanic group exhibited very strong associations for HLA-A~HLA-B (D’= 0.90) and HLA-A~HLA-C (D’= 0.84) haplotypes. Due to the well-characterized recombination hotspot that exists between the HLA-DP and HLA-DQ loci, LD is typically weak (D’ ~0.3) between HLA-DP and other class II genes.

We have conducted an additional analyses shown in Supplementary Table 19 that examines HLA-A to HLA-DQB1 block associations with HLA-DPB1 alleles; in this table it can be observed that some haplotypes extend to HLA-DPB1 with statistically significant chi-squared values however, the D’ values are relatively low. In general HLA-A~HLA-DQB1 blocks show no significant associations with HLA-DPB1 alleles resulting in an overall weak association with the centromeric more distant loci.

Across the majority of populations, which typed the HLA-DQA1 gene, the strength of association was greatest for the HLA-DRB1~HLA-DQA1 haplotype, with D’ ranging from 0.930 in African Americans to maximal LD observed in Arabs. At the HLA-C~HLA-B haplotype D’ was highest in USA Hispanics (D’ = 0.983), API (D’ = 0.972), and the Thai (D’ = 0.978) groups, whilst in the Italian and Greek groups association was strongest for the HLA-DRB1~HLA-DQB1 and HLA-DQA1~HLA-DQB1 haplotypes, respectively. These results reveal that there was no obvious pattern between the strongest LD measures for particular HLA haplotypes and the geographical regions the samples originated from. However, we observed, that African Americans tended to have lower LD values for the majority of the haplotypes compared to the other populations, reflective of the higher gene diversity found in African descent populations.

In general, quantitative estimates of LD values were greater at 4-field than 2-field, irrespective of the distance between the loci, which reflects the greater allelic diversity and the specific allelic haplotypic associations observed for some allele combinations at 4-field. For example, in European Americans HLA-C*08:02:01:01 associates strongly with HLA-B*14:02:01:01 (D’ = 0.98), and HLA-B*14:02:01:02 (D’ = 1), whereas HLA-C*08:02:01:02 associates strongly with HLA-B*14:01:01 (D’ = 0.98) and weakly with HLA-B*14:02:01:01 (D’ = 0.12). When these alleles are reduced to 2-fields, HLA-C*08:02 associates with HLA-B*14:01 (D’ = 0.98) or HLA-B*14:02 (D’ = 0.99) leading to lower global LD values at 2-field (D’ = 0.93) compared to 4-field (D’ = 0.94). Conversely, for common 4-field alleles that associate with many alleles with the same 2-field group global LD values are higher at 2-field compared to 4-field allele data. For instance, the HLA-DPA1~HLA-DPB1 haplotypes consist of predominantly HLA-DPB1*04:01:01:01 that associates tightly with specific low diversity HLA-DPA1 alleles such as HLA-DPA1*01:03:01:02 (D’ = 0.80) and HLA-DPA1*01:03:01:04 (D’ = 0.83). Collapsing these allele names to 2-field results in an equal ratio between HLA-DPA1 and HLA-DPB1 alleles leading to an overall higher LD measure; HLA-DPA1*01:03~HLA-DPB1*04:01, D’ = 0.91.

cALD measures allowed for a more detailed analysis of LD for locus pairs. Noteworthy is the complementary pair of cALD values for the HLA-C~HLA-B haplotype (W_HLA-C/HLA-B and W_HLA-B/HLA-C) observed across all population groups. The W_HLA-C/HLA-B measures were larger than W_HLA-B/HLA-C indicating more diversity of HLA-B alleles compared to HLA-C alleles. Similar cALD patterns were observed for haplotypes HLA-A~HLA-B and HLA-A~HLA-C. At the HLA-DQA1~HLA-DQB1 haplotype, W_{HLA-DQB1/HLA-DQA1} measures were marginally larger than W_{HLA-DQA1/HLA-DQB1} suggesting more allelic diversity of HLA-DQB1 than HLA-DQA1. At the HLA-DPA1~HLA-DPB1 haplotype, the complementary cALD values tended to be similar in European Americans, and Europeans, but W_{HLA-DPA1/HLA-DPB1} was moderately more elevated than W_{HLA-DPB1/HLA-DPA1} in African Americans, USA Hispanics, Mexicans, and Spain. These findings tie into the allele distributions observed in the populations.

3.4.2. Allelic haplotype linkage disequilibrium

Supplementary Tables 4 to 10 shows 2-locus haplotypes (HLA-A~HLA-B, HLA-A~HLA-C, HLA-C~HLA-B, HLA-DRB1~HLA-DQA1, HLA-DRB1~HLA-DQB1, HLA-DQA1~HLA-DQB1, HLA-DPA1~HLA-DPB1) that occurred at least twice for various loci pairs. Distinctive 4-field allele associations were observed across all populations, some were shared across all groups whereas others were ethnic/region-specific.

The HLA-C*07:02:01:03~HLA-B*07:02:01 alleles were either in complete LD or near-complete LD, all at relatively high haplotype frequencies (HF~ 8 – 12%), in populations originating from Europe and USA Hispanics. However this haplotype was also detected in African Americans, Mexicans, APIs, and Indians all with maximal LD but at lower frequencies. In comparison, the alternative 4/3-field HLA-C*07~HLA-B*07 haplotype, HLA-C*07:02:01:01~HLA-B*07:02:01 was more prevalent and LD was stronger in African Americans (HF = 0.034, D’ = 0.73) than European Americans (HF = 0.038, D’ = 0.16); this haplotype was not observed in other groups. Other notable haplotypes common and with significant LD in populations with European ancestry were HLA-C*07:01:01:01~HLA-B*08:01:01:01 (global HF = 0.077) and HLA-C*05:01:01:02~HLA-B*44:02:01:01 (global HF = 0.048). Haplotype HLA-C*03:04:02~HLA-B*15:10:01 was ethnic-specific and found only in African Americans (HF = 0.019, D’ = 0.83, rank 12). A similar example was observed in the API and Thai groups, where haplotype HLA-C*01:02:01~HLA-B*46:01:01 was specific to these groups at maximal LD and occurred at relatively high frequencies of 0.070 and 0.143, respectively.

Due to the lower allelic diversity at the HLA-DQ loci common haplotypes were shared amongst populations, with high LD. Such as haplotypes HLA-DQA1*01:02:01:01SG~HLA-DQB1*06:02:01 (global HF = 0.118, present in all populations except India); HLA-DQA1*02:01:01:01SG ~HLA-DQB1*02:02:01:01 (global HF = 0.093, all populations); HLA-DQA1*03:01:01~HLA-DQB1*03:02:01 (global HF = 0.089, all populations, except India, with the lowest HF seen in African Americans 0.025 highest in European Americans (HF = 0.095). Four haplotypes bearing HLA-DQB1*03:19:01 were observed in African Americans. HLA-DQA1*05:05:01:01SG~HLA-DQB1*03:19:01 was the most frequent in African Americans (HF = 0.041, D’ = 3.9) and Arab’s (HF = 0.039, D’ = 1). This haplotype was observed at lower frequencies in European Americans (HF = 0.001, D’ = 0.66), and Spanish individuals (HF = 0.006, D’ = 0.71), evident of African gene flow in these two latter populations. Noteworthy, is haplotype HLA-DQA1*04:01:02:02~HLA-DQB1*03:19:01 that was only observed in African Americans at a moderate frequency (HF = 0.040, D’ = 0.77).

Interesting allelic haplotypes were observed at the HLA-DP loci. As alluded to earlier, 4-field intronic variants of HLA-DPA1*01:03:01 are relatively common across all groups, and they tend to associate with specific HLA-DPB1 alleles with strong LD. Such as HLA-DPA1*01:03:01:01~HLA-DPB1*02:01:02/HLA-DPB1*02:01:19; HLA-DPA1*01:03:01:02~HLA-DPB1*04:01:01:01/02; HLA-DPA1*01:03:01:03~HLA-DPB1*03:01:01; HLA-DPA1*01:03:01:04~HLA-DPB1*04:01:01:01/02; HLA-DPA1*01:03:01:05~HLA-DPB1*04:02:01:02 are frequent in European Americans, African Americans, USA Hispanic, Mexicans, and Spaniards. However we also observed ethnic specific associations in non-European groups such as HLA-DPA1*01:03:01:02~HLA-DPB1*18:01, HLA-DPA1*02:01:08~HLA-DPB1*01:01:01, HLA-DPA1*03:01~HLA-DPB1*105:01 observed more frequently (HF = 0.07 – 0.12, D’>0.8) in African Americans. In APIs haplotypes HLA-DPA1*02:02:05~HLA-DPB1*05:01:01 (HF = 0.049, D’ = 1), HLA-DPA1*02:02:02~HLA-DPB1*135:01 (HF = 0.037, D’ = 1), HLA-DPA1*02:01:01:02~HLA-DPB1*09:01:01 (HF = 0.024, D’ = 0.5), and HLA-DPA1*02:02:02~HLA-DPB1*02:02 (HF = 0.024, D’ = 1), were common. Whilst in Indians, haplotypes HLA-DPA1*02:01:01:01/02~HLA-DPB1*26: 01:02, and HLA-DPA1*01:03:03~HLA-DPB1*04:01:01:01/02, HLA-DPA1*02:01:02~HLA-DPA1*17:01 appeared to be specific to this group. These haplotypes are prime examples demonstrating the power of NGS to define ethnic specific groups.

3.5. Extended haplotypes comprising of 4–11 loci

Extended haplotypes, that is haplotypes composed of more than 2 loci are detailed in Supplementary Tables 11 to 16 corresponding to A~C~B~DRB1~DQB1, A~C~B~DRB3/4/5~DRB1~DQB1, A~C~B~DRB1~DQB1~DPB1, DRB3/4/5~DRB1~DQA1~DQB1, A~C~B~DRB3/4/5~DRB1~DQA1~DQB1~DPA1~DPB1, and A~C~B~DRB3/4/5~DRB1~DQA1~DQB1 haplotypes respectively. Haplotype counts of 3 estimated by the EM algorithm are generally considered reliable. However, haplotypes counted twice may not be highly accurate but may be informative for populations with small sample sizes. Although haplotypes where n = 2 have been checked for known allele associations we advise the user to err on the side of caution when interpreting or using these haplotypes. Complete extended haplotype datasets, which includes haplotypes with counts less than 2, are available from the 17th IHIW website upon request.

3.5.1. Four locus haplotypes

Silent mutations in HLA-DQ and HLA-DRB3 loci show mutational events that led to present-day haplotype diversity of the HLA-DRB1*03:01:01:01 group. In total nine haplotypes all bearing HLA-DRB1*03:01:01:01SG were estimated (Supplementary Table 14). Of this total, three haplotypes all carried HLA-DRB3*01:01:02:01 but two haplotypes differed at the 4-field for HLA-DQA1*05:01:01, but both carried HLA-DQB1*02:01:01, and one haplotype carried HLA-DQA1*01:02:01:01SG. The rarer haplotype HLA-DRB3*01:01:02:01~HLA-DRB1*03:01:01:01SG~HLA-DQA1*01:02:01:01SG~HLA-DQB1*06:02:01, was detected only in European Americans. Three haplotypes all carried HLA-DQB1*02:01:01 and HLA-DRB3*02:02:01:01 but each carried different intronic HLA-DQA1*05:01 variants; HLA-DQA1*05:01:01:01, HLA-DQA1*05:01:01:02, HLA-DQA1*05:01:01:03. The HLA-DQA1*05:01:01:01 was the most common (global HF = 0.016) across all groups except for USA Hispanics and Arabs where they are absent. These haplotypes frequently exist on extended haplotypes encompassing HLA-C*05:01:01:01 and HLA-B*18:01:01:01 alleles. In comparison two haplotypes all had HLA-DRB3*02:02:01:02 and HLA-DQB1*02:01:01 but either carried HLA-DQA1*05:01:01:01 or HLA-DQA1*05:01:01:02. We see a similar pattern of high-level diversity for the HLA-DRB1*01 bearing haplotypes, where both antigenic and intronic differences at HLA-DQA1 and HLA-DQB1 account for distinct haplotypes.

It is noteworthy that there are different eight variations of the HLA-DRB1*07:01:01:01SG~ HLA-DQA1*02:01:01:01SG haplotype each bearing different HLA-DQB1 and HLA-DRB4 alleles. The HLA-DRB4~HLA-DRB1*07:01:01:01SG may bear HLA-DQA1 alleles HLA-DQA1*02:01:01 (most common across all groups) or HLA-DQA1*03:03:01:01 (relatively infrequent and only found in African Americans), and HLA-DQB1 alleles HLA-DQB1*02:02:01:01, HLA-DQB1*02:02:01:02, HLA-DQB1*03:01:01:03 and HLA-DQB1*03:03:02:01. The HLA-C*16:01:01:01~HLA-B*44:03:01:01 block commonly exists on the HLA-DRB4*01:01:01:01~HLA-DRB1*07:01:01:01SG~HLA-DQA1*02:01:01:01SG~ HLA-DQB1*02:02:01:01 haplotype. HLA-DRB4*01:03:01:01~HLA-DRB1*07:01:01:01SG~ HLA-DQA1*02:01:01:01SG~HLA-DQB1*02:02:01:01 extended haplotypes bear either HLA-C*06:02:01:01~HLA-B*13:02:01, HLA-C*06:02:01:02~HLA-B*50:01:01, or HLA-C*16:01:01:01~HLA-B*44:03:01:01. Also HLA-DRB4*01:03:01:02N~HLA-DRB1*07:01:01:01SG~HLA-DQA1*02:01:01:01SG~HLA-DQB1*03:03:02:01 exists on haplotypes bearing HLA-C*06:02:01:01~HLA-B*57:01:01.

Asians have distinctive HLA-DRB1*15 haplotype blocks compared to other groups. They typically carry HLA-DRB5*01:02~HLA-DRB1*15:02:01:01 with different HLA-DQ alleles differing at the antigenic level; HLA-DQA1*01:03:01:01, HLA-DQA1*01:02:01:01SG, HLA-DQA1*01:01:01:02SG, HLA-DQB1*06:01:01, HLA-DQB1*05:02:01, and HLA-DQB1*05:01:01:01.

3.5.2. Nine locus haplotypes based on eleven locus genotyping data

We estimated extended haplotypes both in the presence (Supplementary Table 15) and the absence of HLA-DP loci (Supplementary Table 16) due to the weaker association of HLA-DP alleles with alleles at other HLA loci. Overall, the most frequent extended haplotypes predicted by the EM-algorithm reflect the allele distribution found in individual groups and are essentially composed of the most frequent 2-locus haplotypes. Extended ancestral haplotypes mediated by strong LD between distant class I and class II alleles have been well-described in the literature [55–57]. The most common 9 loci haplotype was HLA-A*01:01:01~HLA-C*07:01:01:01~B*08:01:01:01~HLA-DRB3*01:01:02:01~HLA-DRB1*03:01:01:01SG~HLA-DQA1*05:01:01:02~HLA-DQB1*02:01:01~HLA-DPA1*01:03:01:02~HLA-DPB1*04:01:01:01/HLA-DPB1*04:01:01:02 (global HF = 0.021) which corresponds to the most frequent ancestral haplotype (AH8.1) observed in ~5–12% of Northern European populations [56,57]. Interestingly, this haplotype was the most common in European Americans (HF = 0.028) but not in the European population where it was ranked number 3 and haplotype HLA-A*03:01:01~HLA-C*07:02:01:03~HLA-B*07:02:01~HLA-DRB5*01:01:01~HLA-DRB1*15:01:01:01SG~HLA-DQA1*01:02:01:01SG~HLA-DQB1*06:02:01~HLA-DPA1*01:03:01:02~HLA-DPB1*04:01:01:01/HLA-DPB1*04:01:01:02 was the most frequent (HF = 0.024); the low resolution equivalent of this haplotype was the 4^th most frequent in a Norwegian population. The discrepancy in ranking may be due to the small sample size of the Europeans relative to the European Americans.

In African Americans, the most prevalent haplotype was HLA-A*68:01:01:02~HLA-C*06:02:01:01~HLA-B*58:02:01~HLA-DRB3*02:02:01:02~HLA-DRB1*12:01:01:03~HLA-DQA1*01:05:01~HLA-DQB1*05:01:01:02~HLA-DPA1*01:03:01:02~HLA-DPB1*18:01 (HF = 0.011); this haplotype was not observed in any other population. The haplotype consists of alleles that are prevalent to Africans such as HLA-B*58:02:01 and HLA-DPB1*18:01, also the 2-locus haplotype HLA-DQA1*01:05:01~HLA-DQB1*05:01:01:02 is most frequent in African groups.

Although 9 loci were estimated in all populations, which typed these loci, the numbers were generally too small in other groups to make definite conclusions, for example in USA Hispanics the two most frequent haplotypes were found at only 4 copies corresponding to a frequency of 0.042.

4. Discussion

In this report, we detailed the major findings of the collaborative efforts initiated under the auspices of the 17^th IHIW unrelated HLA diversity component. This study involved many laboratories worldwide and focused on the collection of diverse populations, for detailed characterization of HLA allelic variants determined using high-resolution NGS methods. The HLA typing data was then used to infer haplotype frequencies, and together with allele frequency data were utilized to conduct a thorough analysis of HLA diversity in worldwide populations. In addition, we quantitated LD between loci and alleles in order to gain fresh insights into the demographic history of populations. Due to the power of NGS, the unrelated population data submitted to the 17th IHIW has extended and provided new HLA population data compared to previous workshops [23,26,27]. In this study we showed that there are distinct differences in 4-field allele, haplotype, and LD patterns in populations originating from different geographical regions of the world.

In regards to allele frequencies, although a relatively high number of class I and class II alleles were shared among populations they differed in their frequencies. For instance, HLA-A*02:01:01:01 was found in all groups, but on average was found at almost twice the frequency in Europeans (32%) compared to African Americans (15%) and USA Hispanics (13%), and the lowest frequencies were found in groups from Asia. Another example is given by HLA-B*08:01:01:01 which was moderately high in populations of Eurasian ancestry (~10%), lower in African Americans (3.5%), and even lower in the Asian groups (~1%). Although admixture was evident in the USA populations unique alleles were also observed in these groups. These trends are also observed for some alleles at the class II loci, such as HLA-DRB1*12:01:01:03 which is relatively common in Eurasian ancestry groups, in contrast its 4-field counterpart HLA-DRB1*12:01:01:01 is more prevalent in Asia [36]. Alleles that were detected exclusively in a single population were also observed. We can refer to such alleles as ‘region specific’ eloquently described by Meyer and colleagues [23], since they are present in populations derived from distinct geographical regions. In our study, region specific alleles (RSA) accounted for a remarkable 42% of the total unique alleles identified, and the majority of these alleles were found at relatively low frequencies, and tended to be less abundant at the least polymorphic loci, in agreement with the Meyer study. Overall, the largest proportion of RSA’s was found at HLA-DQB1 (51%) and lowest found at HLA-DPA1 (4.4%). As expected the largest population examined, that is European Americans had the most RSA and rare alleles; rare alleles are indicative of a rapidly expanding population. A prime example is demonstrated by HLA-A*02:60:01 and HLA-B*07:12 which were found exclusively, both at 0.13% allele frequency, in African Americans. It is noteworthy that HLA-B*07:12 has also been reported in Ashkenazi, Ethiopia, Libya, and Morocco Jews albeit at lower frequencies (~0.001%) according to www.allelefrequencies.net. Due to the slightly higher frequency and the fact that the African American individual who carried HLA-B*07:12 also carried HLA-C*15:05:02, which is a well-known to be frequent in African populations, we speculate that HLA-B*07:12 is an African allele and was detected in the Jewish groups due to African gene flow mediated by migration. Surprisingly, an RSA HLA-A*02:11:01 was found at a relatively high frequency in the Indian cohort (AF = 7.1%), this allele is in significant strong LD with HLA-B*40:06:01:02; HLA-B*40:06:01:02 occurs at very low frequencies in Eurasian population but is elevated in Indians at 8.6%. The majority of individuals in the Indian sample were from Southern Indian Andhra Pradesh population, and recently HLA-A*02:11:01 has recently been reported in this population [58]. Although, in this study we define HLA-A*02:11:01 as region specific this allele has also been observed in aboriginal Indians from South America at low frequencies [59]. The limited sample size of some of the populations included in this study has reduced the number of RSA and low-frequency alleles detected.

We observed several population specific alleles that differed at the 4-field resolution. For the HLA-A*24:02:01 group, 7 intronic variants were identified in the entire sample, however HLA-A*24:02:01:07 was found solely in African Americans at a low frequency (AF = 0.13%), whilst HLA-A*24:02:01:06 was observed in USA Hispanics (AF = 1.8%); HLA-A*24:02:01:01 was detected in all other groups at low or intermediate frequency. Similarly, at the HLA-30:02:01 group three 4-field variants were found in our dataset, but only HLA-A*30:02:01:03 was detected in African Americans at 1.1%. These findings illustrate two things; (i) the power of NGS for capturing variants in intron and untranslated regions resulting in more fine-scale discrimination between populations, (ii) the different natural selection pressures acting on intronic segments in alleles from distant geographical regions. Since our data is fairly current, it may be too preliminary to speculate why the specific 4-field alleles examples described above are region specific. However, it is well known that introns may harbor functional polymorphisms that can directly influence gene expression, affect splicing, and may confer risk or protection to disease [6,60]; specific 4-field variants may have developed due to the different pathogen environments.

The large proportion of RSA identified in this dataset is likely to have influenced the deviations from HWE found at several loci in many populations. In addition, HWE deviations can also be explained by recent genetic admixture, inaccurate population sampling, genotyping errors, and natural selection in favor of an excess of HLA heterozygotes which could confer resistance to pathogens and cancers. For this reason, the population samples out of HWE may not be considered as true representatives of the population, and the data should be used with caution.

This study also identified several common alleles with identical protein sequences differing only by intronic substitutions that were found in distinct haplotypes suggesting that their origins were distant or independent. This observation together with the previously well-described protein sequence variability suggests that both diversification and convergent evolution have been at play in shaping the present-day variation of the HLA region.

The variation in allele frequencies leads to diverse haplotype frequencies among populations, and in turn haplotype variation gives rise to a complex pattern of LD across the MHC. LD across the MHC is known to be strong and extensive, creating extended ancestral haplotypes across the entire region. A comparison of LD distribution across populations revealed some interesting observations. Due to the relatively high level of genetic heterogeneity and ancient history of African populations, LD in African descent populations is expected to be less than observed in European as well as other populations and also less structured [54,61]. This means that LD from diverse groups usually cannot be extrapolated to other populations, highlighting the importance of characterizing LD patterns in different populations. As expected, we observed lower levels of LD in the African American population compared to others. Although Africans Americans are often used as a proxy for an African group, several genetic studies have shown that they harbor ancestry from many parts of West Africa as well as 10–25% of European admixture depending on the geographical location in the US. Unfortunately Africa is not well represented in this study, the Arab group from North Africa (the United Emirates of Arab) is too small and is known to be highly admixed with European, African, and indigenous groups to make any concrete interpretations about short and long range LD patterns. In order to better capture HLA diversity on a more detailed level in the African continent, the next 18th Workshop should focus on recruitment of population samples from several widespread African countries.

The higher levels of LD observed across the European descent groups, is indicative of low levels of recombination. Also the long-range LD with high allele frequency, observed with the so-called 8.1 or Cox haplotype, is indicative of recent positive selection, because if an allele has preserved ancestral relationships with neighboring alleles, this implies that the population is relatively young.

Conclusion

The introduction of next-generation sequencing (NGS) based genotyping technologies created valuable opportunities for collaboration and learning for the HLA and Immunogenetics (H&I) community. With a clear goal of advancing the fields of H&I research through the application of NGS technologies, the 17th IHIW created the foundation and structure for a centralized database system the tools developed for the examination of HLA diversity in multiple world populations. The different HLA profiles in unique populations provide valuable information for studying the molecular evolution of HLA and also for association analysis with immune and autoimmune diseases. Our analyses was somewhat hindered by the relatively small sample size for some groups such as Asians and Thais. As a consequence allele and haplotype frequencies may not reflect those observed in larger or entire populations. Larger samples sizes are required for adequate analyses particularly for the more heterogenetic groups such as USA Hispanics, African Americans and Arabs. The 17^th UPHD study highlights the importance of high resolution typing, including intronic and flanking segments for analyzing population differences. Our data provides valuable 2-locus haplotype maximum resolution data that are useful and more informative than allele frequencies to determine population differences. Our study presents analyses of associations of alleles at the under-studied HLA-DRB3, HLA-DRB4, HLA-DRB5, HLA-DQA1, and HLA-DPA1 loci with alleles from flanking loci and their presence in extended haplotypes and corresponding blocks. In spite of the sample size limitations of this study and the lower abundance of non-admixed populations, the results provided by this study illustrate the even larger haplotypic diversity defined by variations in the non-coding regions. These observations may lend information leading to plan and conduct new studies focused in further understanding the evolution of the human MHC. In addition, the information provided here may help with the design of novel disease association studies in selected populations that may allow for finer mapping of the disease determining HLA associated factors. In this regard the rich variation of associations between alleles at HLA-DRB1 and HLA-DRB5 loci in some Asian populations allows us to speculate that the primary and secondary roles of HLA-DRB1 and HLA-DRB5 alleles may be dissected in diseases associated with haplotypes bearing alleles at the HLA-DRB5 locus. The present study identified fully extended haplotypes that included alleles at loci not evaluated extensively in the past. These findings taken together with the inclusion of distinct populations may provide significant information about the chances of finding closely HLA matched and/or antibody compatible unrelated donors for sensitized HSCT patients with unique, understudied or admixed ancestry. With the addition of new reference sequences and alleles to the IMGT/HLA Database, we anticipate that more accurate allele calling will be achieved. It should be noted that after the 17^th IHIW the sequence intervals considered for naming of HLA alleles has been expanded for the class I loci resulting in the addition of new names very often with the preservation of the initial allele name; therefore, some of the allele assignments could change with some possible changes in the associations described in the present study. The impact of improvements in HLA assignments and changes in HLA Nomenclature will be examined in some projects of the 18^th IHIW. The tables presented in this manuscript are available in the 17^th IHIW website (http://17ihiw.org/17th-ihiw-ngs-hla-data/).

Figure 3.

Global linkage disequilibrium values of D’ and W_n class I haplotypes by population.

Abbreviations:

HLA

Human leukocyte antigen

UPHD

Unrelated Population HLA Diversity project

AF

Allele frequency

HF

Haplotype frequency

F

homozygosity

LD

linkage disequilibrium

cALD

conditional asymmetric linkage disequilibrium

CEH

Conserved extended haplotypes