Abstract
The complex pathogenesis of rheumatoid arthritis (RA) is not fully understood, with few studies exploring the genomic contribution to RA in patients from Africa. We report a genome-wide association study (GWAS) of South-Eastern Bantu-Speaking South Africans (SEBSSAs) with seropositive RA (n = 531) and population controls (n = 2653). Association testing was performed using PLINK (logistic regression assuming an additive model) with sex, age, smoking and the first three principal components as covariates. The strong association with the Human Leukocyte Antigen (HLA) region, indexed by rs602457 (near HLA-DRB1), was replicated. An additional independent signal in the HLA region represented by the lead SNP rs2523593 (near the HLA-B gene; Conditional P-value = 6.4 × 10−10) was detected. Although none of the non-HLA signals reached genome-wide significance (P < 5 × 10−8), 17 genomic regions showed suggestive association (P < 5 × 10−6). The GWAS replicated two known non-HLA associations with MMEL1 (rs2843401) and ANKRD55 (rs7731626) at a threshold of P < 5 × 10−3 providing, for the first time, evidence for replication of non-HLA signals for RA in sub-Saharan African populations. Meta-analysis with summary statistics from an African-American cohort (CLEAR study) replicated three additional non-HLA signals (rs11571302, rs2558210 and rs2422345 around KRT18P39-NPM1P33, CTLA4-ICOS and AL645568.1, respectively). Analysis based on genomic regions (200 kb windows) further replicated previously reported non-HLA signals around PADI4, CD28 and LIMK1. Although allele frequencies were overall strongly correlated between the SEBSSA and the CLEAR cohort, we observed some differences in effect size estimates for associated loci. The study highlights the need for conducting larger association studies across diverse African populations to inform precision medicine-based approaches for RA in Africa.
Introduction
Rheumatoid arthritis (RA) is a chronic systemic inflammatory disease that principally affects peripheral synovial joints (1). Although it was previously thought to be uncommon and milder in patients of sub-Saharan African (SSAf) ancestry, there is now a growing body of evidence that links the disease to impaired physical function and reduced health-related quality of life, especially in urban African populations (2,3). The interplay of environmental factors (e.g. smoking) and genetic factors acts as a trigger for immune-mediated chronic inflammation in seropositive RA (1,4,5). The class II Human Leukocyte Antigen (HLA) region is consistently the strongest genetic locus associated with RA and accounts for over 30% of the heritability in most populations studied, including SSAf populations (6–8). Studies in black South Africans have shown that HLA-DRB1 alleles that carry the ‘shared epitope’ (SE) are associated with RA (9). More recently, fine mapping analyses have shown that individual amino acids at positions 11 and 13 of the third hypervariable region of the beta chain conferred the strongest risk for RA in black South Africans and African-Americans (AAs) (10–12).
Genome-wide association studies (GWASs) performed globally have revealed significant inter-population variation. For example, PTPN22 is the major non-HLA genetic risk factor in patients of European ancestry, but is not associated with RA in East Asians, AA or SSAf populations, with a near total absence of the R620W variant in those populations (13–17). PADI4 confers risk for developing RA in Asian and in some European ancestry patients (18,19). Furthermore, data from a recent study (13) suggest that the RA risk loci with larger effect sizes in one population, such as ILF3, TYK2 and IL20RB, are less likely to replicate in other populations owing to differences in allele frequencies across populations. Given the limited understanding of genetic susceptibility for RA in populations from Africa (12,14), we undertook a genome-wide case–control association study of South-Eastern Bantu-Speaking South Africans (SEBSSAs) (previously and commonly referred to as black South Africans) with seropositive RA.
Results
Patient demographics
Among RA patients, 86% were female, in contrast to 56% in the control group. Mean age (±SD) was similar in cases (46.6 ± 15.2 years) and controls (50.2 ± 6.8 years). More controls than patients were smokers (37.5% vs. 18.6%). Similar to observations in other countries (20), South African females are disproportionately affected by RA. Since South African females have a much lower rate of smoking (21), the difference in sex composition between cases and controls has likely contributed to the higher rate of smoking among controls. All patients were seropositive for anti-citrullinated peptide antibodies (ACPA) and 98% for rheumatoid factor (RF).
SEBSSA GWAS
We performed a GWAS of 531 RA cases and 2653 control SEBSSA participants (Fig. 1A, Supplementary Material, Fig. S1). Due to population sub-structure observed in principal component analysis (PCA) (Supplementary Material, Fig. S2), the first three principal components (PCs) were used as covariates along with age, sex and smoking for association testing. The HLA region emerged as the key signal with rs602457 (near HLA-DRB1) as the lead SNP (Fig. 1A; Supplementary Material, Fig. S3). A conditional analysis using Genome-wide Complex Trait Analysis (GCTA) (22) in addition to verifying this lead signal identified two independent signals rs2523593 (near HLA-B, Conditional P-value 6.4 × 10−10) and rs29001652 (near HLA-DRA, Conditional P-value 5.4 × 10−12) within the region (Fig. 1B–E; Supplementary Material, Table S1).
Figure 1.
Summary of associations detected in the GWAS and the meta-analysis. (A) Miami plot showing association signals from the SEBSSA GWAS (upward facing) and SEBSSA-CLEAR meta-analysis (downward facing). Signals with P-value<10−10 have been shown as P-value = 10−10 to enhance the visibility of suggestive signals. Only SNPs that were common to the two studies are shown. The red horizontal line corresponds to the genome-wide significance threshold (P-value = 5 × 10−8) and the orange dots represent SNPs below the suggestive threshold (P-value = 5 × 10−6). (B) Regional association plot for HLA in the SEBSSA study. Three independent signals identified using conditional analysis are shown in blue. Locuszoom plots showing LD and association statistics in the +/−200 kb region around the lead SNPs corresponding to the three independent signals (C) rs2523593 (D) rs29001652 and (E) rs602457. The LD estimates were based on the SEBSSA dataset.
Given the complexity of the HLA region, it was important to assess whether these independent signals are due to the linkage disequilibrium (LD) architecture of the imputation panel used. Therefore, we imputed the data using the Michigan multi-ethnic HLA reference panel (23) and performed an additional association and COJO analysis for the HLA region based on this dataset. The analysis identified a different SNP [6:32553343 (SNPS_DRB1_5239_32552281_intron1_AG)] near HLA-DRB1 (not present in our original imputed dataset, but very close to the lead rs602457) as the lead SNP for the region (Supplementary Material, Fig. S4, Supplementary Material, Table S1). Conditional analysis on this dataset confirmed rs2523593 as an independent association signal within the region (Supplementary Material, Table S1).
Seventeen independent genomic regions outside the HLA/chromosome 6 region showed suggestive associations by Functional Mapping and Annotation tool (FUMA) (Table 1, Supplementary Material, Table S2, Supplementary Material, Fig. S5). The most notable of these was a set of signals on chromosome 4, with lead SNP rs75806510 (P-value = 2.5 × 10−07, Beta = 0.54) that mapped to an intron of a long non-coding RNA (RP11-79E3.3). The variant allele was present at a frequency of 7% in African populations but was absent in European, South Asian and East Asian populations from the 1000 Genomes dataset (24). In accordance with its absence in Eurasian populations, we did not detect any previous associations of this SNP with RA in the GWAS literature. No previous associations for RA were detected in the surrounding genomic region (+/−100 kb) making this signal novel and perhaps unique to African ancestry populations. Other suggestive associations included signals in or near the DDR2, ARL2BPP7, CHMP3, RGS10 and PREX1 genes.
Table 1.
Suggestive non-HLA associations with rheumatoid arthritis (RA) in the SEBSSA GWAS
| GLa | rsID | Chr | P-value | EAF | Start | End | nGS | Lead SNPs | Genes |
|---|---|---|---|---|---|---|---|---|---|
| 1 | rs61782488 | 1 | 1.60E-06 | 0.135 | 101 773 724 | 101 839 211 | 16 | rs61782488 | RP11-157 N3.1 |
| 2 | rs11580729 | 1 | 3.92E-06 | 0.209 | 162 593 472 | 162 598 144 | 4 | rs11580729 | UAP1; DDR2 |
| 3 | rs2674826 | 2 | 2.40E-06 | 0.282 | 86 930 966 | 87 003 846 | 3 | rs2674826 | CHMP3; RMND5A |
| 4 | rs13420653 | 2 | 1.88E-06 | 0.311 | 221 696 579 | 221 712 773 | 8 | rs13420653 | - |
| 5 | rs79495193 | 3 | 3.07E-06 | 0.415 | 120 243 465 | 120 255 261 | 7 | rs79495193 | RP11-359H3.1 |
| 6 | rs1021661 | 3 | 8.46E-07 | 0.238 | 192 686 115 | 192 754 872 | 17 | rs1021661 | - |
| 7 | rs75806510 | 4 | 2.50E-07 | 0.164 | 33 644 214 | 34 113 947 | 124 | rs75806510 | RP11-79E3.3 |
| 8 | rs12656393 | 5 | 4.98E-06 | 0.242 | 174 692 575 | 174 737 779 | 8 | rs12656393 | ARL2BPP6 |
| 9 | rs115216679 | 5 | 2.39E-06 | 0.174 | 180 615 059 | 180 615 059 | 1 | rs115216679 | CTC-338 M12.2 |
| 10 | rs1013012 | 8 | 2.11E-06 | 0.076 | 1 301 287 | 1 360 948 | 1 | rs1013012 | |
| 11 | rs3918975 | 9 | 1.79E-07 | 0.159 | 104 792 890 | 104 792 890 | 1 | rs3918975 | ARL2BPP7 |
| 12 | rs12219225 | 10 | 2.85E-06 | 0.534 | 121 278 224 | 121 280 683 | 4 | rs12219225 | RGS10 |
| 13 | rs116603585 | 11 | 1.25E-06 | 0.084 | 57 431 047 | 57 812 949 | 4 | rs116603585 | OR5AZ1P, RP11-659P15.1 |
| 14 | rs61255797 | 14 | 4.47E-06 | 0.271 | 25 966 990 | 25 966 990 | 1 | rs61255797 | - |
| 15 | rs60238262 | 14 | 1.83E-06 | 0.163 | 62 325 505 | 62 326 913 | 4 | rs60238262 | CTD-2277 K2.1 |
| 16 | rs12964352 | 18 | 6.16E-07 | 0.231 | 77 388 091 | 77 407 662 | 8 | rs12964352 | RP11-567 M16.3 |
| 17 | rs113014384 | 20 | 3.18E-06 | 0.151 | 47 432 840 | 47 473 162 | 2 | rs113014384 | PREX1 |
GL: index of genomic rick loci.
rsID: rsID of the top lead SNP based on dbSNP build 146.
Chr: chromosome of top lead SNP
P-value: P-value of top lead SNP
EAF: effect allele frequency in the SEBSSA GWAS
Start: start position of the locus.
End: end position of the locus.
nGS: the number of unique GWAS-tagged candidate SNPs in the genomic locus which are available in the GWAS summary statistics input file.
LeadSNPs: rsID of lead SNPs in the genomic locus. For this the signals were first separated into independent blocks/loci based on LD. Each block/locus is then partitioned into lead SNPs.
Genes: genes detected within/around the genomic region. Intergenic regions are shown with a ‘-’
aClassification of signals into a genomic locus was performed using FUMA using African LD from 1000 Genomes project.
Meta-analysis with CLEAR cohort
The meta-analysis of the SEBSSA and Consortium for the Longitudinal Evaluation of African-Americans with Early Rheumatoid Arthritis (CLEAR) cohort summary statistics (Fig. 1; Supplementary Material, Fig. S6) supported the presence of at least two independent signals in the HLA region (corresponding to rs602457 and rs2523593, but with different lead SNPs) (Supplementary Material, Fig. S7). However, HLA is a very difficult region to characterize due to the complex LD architecture and the loss of SNPs due to consideration of only the SNPs that were common to both studies in the meta-analysis could have impacted the accuracy of LD estimates. As the imputation panels used were also different for SEBSSA and CLEAR, there was limited scope to perform a reliable conditional analysis to explore this further.
The meta-analysis did not detect any non-HLA associations at the genome-wide significance threshold. The suggestive associations identified signals near NEDD4, ARL2BPP7 and PRKCQ genes (Fig. 1A, Supplementary Material, Table S3). Although, we observed some support and beta direction consistency for the suggestive association on chromosome 4 (rs75806510) in the meta-analysis (SEBSSA P-value = 2.5 × 10−7, Beta = 0.545; CLEAR P-value = 0.09, Beta = 0.291; Meta-analysis P-value = 1.572 × 10−7, Beta = 0.456), the other suggestive signals detected in the SEBSSA GWAS did not receive any boost in signal strength in the meta-analysis. Importantly, some of the suggestive signals were not included in the meta-analysis because of their absence in the CLEAR dataset.
Replication of known signals
To test for replication of known association signals, we noted the P-values of over 300 previous RA-associated signals from GWAS Catalog (25) in the SEBSSA GWAS results. Among the previously characterized GWAS signals, nine HLA SNPs crossed the genome-wide significance threshold in the SEBSSA cohort (Supplementary Material, Table S4). Additional five SNPs were observed to show replication at a score-test defined P-value cut-off of 5 × 10−3 (see Methods for details). Among the non-HLA signals, a SNP each in the MMEL1 (rs2843401) and ANKRD55 (rs7731626) genes showed replication in the SEBSSA cohort. This to our knowledge is the first report of replication of a non-HLA signal detected in European, East Asian and trans-ancestry (26) GWASs in a SSAf dataset. The meta-analysis replicated three additional non-HLA signals, two on chromosome 2 (near the KRT18P39 and CTLA4 genes) and a signal on chromosome 1 (AL645568.1 gene) (Table 2).
Table 2.
Previously reported RA associations that show exact replication in the SEBSSA-CLEAR meta-analysis
| Chr | Pos | rsID | META | SEBSSA | CLEAR | Mapped Genes | |||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| BETA | P-value | EAF | BETA | P-value | EAF | BETA | P-value | ||||
| 6 | 32 582 650 | rs3104413 | 0.888 | 5.60E-35 | 0.291 | 1.009 | 2.59E-27 | 0.115 | 0.746 | 2.58E-10 | |
| 6 | 32 444 198 | rs12194148 | 0.767 | 1.23E-33 | 0.458 | 0.873 | 1.17E-28 | 0.303 | 0.627 | 1.43E-07 | |
| 6 | 32 428 772 | rs9268839 | 0.751 | 1.79E-32 | 0.459 | 0.863 | 8.79E-28 | 0.295 | 0.608 | 2.66E-07 | |
| 6 | 32 577 380 | rs660895 | 0.687 | 9.52E-26 | 0.321 | 0.772 | 2.27E-19 | 0.158 | 0.586 | 1.05E-08 | |
| 6 | 32 669 955 | rs9275406 | 0.564 | 1.08E-22 | 0.504 | 0.704 | 2.14E-19 | 0.326 | 0.416 | 5.17E-06 | |
| 6 | 32 670 978 | rs9275428 | 0.564 | 1.08E-22 | 0.504 | 0.704 | 2.14E-19 | 0.326 | 0.416 | 5.17E-06 | |
| 6 | 32 663 851 | rs6457617 | -0.549 | 3.61E-21 | 0.353 | −0.552 | 1.20E-12 | 0.526 | −0.545 | 2.24E-10 | |
| 6 | 32 451 788 | rs9269234 | −0.465 | 1.52E-10 | 0.261 | −0.490 | 4.36E-09 | 0.335 | −0.393 | 0.00534 | |
| 6 | 32 375 973 | rs3763309 | 0.395 | 2.30E-06 | 0.059 | 0.142 | 0.371692 | 0.116 | 0.629 | 3.17E-07 | |
| 6 | 32 671 103 | rs13192471 | 0.277 | 4.48E-06 | 0.355 | 0.362 | 5.38E-06 | 0.243 | 0.172 | 8.61E-02 | |
| 6 | 32 218 989 | rs9296015 | 0.215 | 2.29E-04 | 0.216 | 0.395 | 1.8E-05 | 0.243 | 0.029 | 7.76E-01 | |
| 2 | 204 636 190 | rs2558210 | 0.182 | 1.58E-03 | 0.489 | 0.199 | 7.2E-03 | 0.597 | 0.157 | 7.24E-02 | KRT18P39, NPM1P33 |
| 1 | 2 528 133 | rs2843401 | 0.174 | 1.71E-03 | 0.497 | 0.257 | 5.3E-04 | 0.464 | 0.070 | 4.72E-01 | MMEL1 |
| 1 | 173 337 747 | rs2422345 | -0.168 | 3.11E-03 | 0.435 | -0.118 | 1.1E-01 | 0.558 | -0.233 | 5.8E-03 | AL645568.1 |
| 2 | 204 742 934 | rs11571302 | -0.177 | 4.21E-03 | 0.383 | -0.110 | 1.5E-01 | 0.346 | -0.262 | 4.6E-03 | CTLA4, ICOS |
Only SNPs that were included in both the datasets are shown (for example rs7731626 that was not included in the CLEAR dataset is not shown). P < 0.005 in non-HLA loci are shown in bold. Non-HLA genes are named.
Chr, Pos and rsID: chromosome, position and rsID of the previously reported SNP.
META BETA, P-value: effect size and P-value for the meta-analysis.
SEBSSA EAF: frequency in the SEBSSA dataset
SEBSSA BETA, P-value: effect size and P-value for the SEBSSA GWAS
CLEAR EAF: frequency in the CLEAR dataset
CLEAR BETA, P-value: effect size and P-value for the CLEAR GWAS.
We also performed regional replication for non-HLA signals in both the GWAS and meta-analysis results (Table 3, Supplementary Material, Table S5). For this, 100 kb windows on either side of the known GWAS signals were scanned and SNPs showing a P-value lower than the score-test defined threshold of 2 × 10−4 were considered to be replicated. A signal near LOC107984408 was observed to be replicated in the SEBSSA GWAS. This approach replicated a well-known association near the PADI4 gene in the meta-analysis. The SNP rs4262594 from this region showed P-values <0.005 in both SEBSSA and CLEAR GWASs and a suggestive level P-value (P-value<3.4 × 10−6) in the meta-analysis. Signals near CD28, LIMK1, ZNF679, DNASE1L3 and LINC02098-ETS1 genes were also found to be replicated in the meta-analysis at this threshold.
Table 3.
Previously reported non-HLA RA associations that show regional replication (+/− 100 kb) in the SEBSSA-CLEAR meta-analysis. Further details in Supplementary Material, Table S5
| Chr | Pos | rsID | Mapped gene | Replication | P-value | Distance (kb) |
|---|---|---|---|---|---|---|
| 1 | 17 655 407 | rs1748041 | PADI4 | rs10888012 | 4.51E-06 | 59.596 |
| 2 | 100 760 172 | rs11123811 | AFF3 | rs11692867 | 1.46E-04 | 0.695 |
| 2 | 204 636 190 | rs2558210 | KRT18P39—NPM1P33 | rs12614091 | 9.59E-05 | 3.329 |
| 3 | 58 183 636 | rs35677470 | DNASE1L3 | rs60415364 | 5.52E-05 | −14.968 |
| 7 | 63 726 645 | rs1830035 | ZNF679 | rs61456016 | 6.21E-05 | 4.551 |
| 7 | 73 537 902 | rs193107685 | LIMK1—EIF4H | 7:73588782 | 1.35E-04 | −50.88 |
| 9 | 34 710 338 | rs11574914 | AL162231.5 | rs12348523 | 1.12E-04 | −1.297 |
| 10 | 6 517 167 | rs502919 | PRKCQ | rs11258964 | 1.11E-04 | 16.088 |
| 11 | 128 156 314 | rs12795702 | LINC02098—ETS1 | rs7941730 | 1.90E-05 | −54.191 |
| 16 | 23 871 206 | rs149041927 | PRKCB | rs4018633 | 1.86E-04 | 54.794 |
| 19 | 17 434 093 | rs77331626 | ANO8, DDA1 | rs8110443 | 1.42E-04 | −74.583 |
Chr, Pos and rsID: chromosome, position and rsID of the previously reported SNP
Mapped gene: genes/nearby gene based on GWAS catalog
Replication: ID of the closest SNP to the original signal that showed the P-value below the replication threshold in the meta-analysis (P-value<2E-04).
P-value: P-value for association in meta-analysis
Distance: distance between the previously reported and the replication signals (kb).
Comparisons of signals between CLEAR and SEBSSA cohorts
The PCA plot showing our cohort along with representative African and AA populations from the 1000 Genomes Project dataset highlights clear genetic differences between our cohort and AA (represented here by the ASW) populations (Supplementary Material, Fig. S8). To analyze the extent to which the level of intrinsic differences between the two populations impact association study-related estimates, we compared effect allele frequencies (EAFs) and effect sizes for known RA-associated SNPs that were at least nominally replicated (P < 0.05 for non-HLA and P < 0.005 for HLA regions) in the meta-analysis (Fig. 2). We observed an overall high concordance between EAFs for most of the SNPs, with exceptions such as rs12525220 (SEBSSA MAF 0.14; CLEAR MAF 0.07), rs7765379 (SEBSSA MAF 0.14; CLEAR MAF 0.09) and rs9296015 (SEBSSA MAF 0.16; CLEAR MAF 0.24) that showed 1.5-fold or higher frequency difference between the two cohorts. The effect sizes on the other hand showed noticeable differences in the two cohorts. These were more pronounced for non-HLA signals compared with HLA signals. Similar trends in effect size and allele frequency differences were observed in the suggestive associations detected in the SEBSSA cohort (Supplementary Material, Fig. S9).
Figure 2.
Comparison of effect sizes and EAFs in SEBSSA and CLEAR cohorts. Known RA-associated SNPs (from GWAS Catalog) that show a P-value<0.05 for non-HLA and P-value<0.005 for HLA in the SEBSSA summary statistics are included. (A) Effect size comparisons of HLA signals. (B) Effect size comparisons of non-HLA signals. (C) EAF comparisons of HLA signals (D). EAF comparisons of non-HLA signals. A more comprehensive set of SNPs are shown in Supplementary Material, Figure S9.
Discussion
RA is a complex multifactorial disease of the joints whose pathogenesis remains poorly understood. Although GWASs have identified over a hundred RA susceptibility SNPs in European and Asian ancestry populations, there is a dearth of genetic data on African populations. Our understanding of the genetics of RA in African ancestry populations so far has been primarily driven by studies conducted on AA. We report, to our knowledge, the first GWAS study of RA in a continental SSAf population.
It is noteworthy that there is considerable population structure across African populations which is likely to influence disease associations. The Southern African Bantu-speakers are the only African population to harbor a large proportion of genetic ancestry from the Khoe and San people, who represent the earliest lineage that diverged from all other modern human lineages (27). The level of gene flow varies with geography and ethnolinguistic group, from almost none in some groups to > 20% in others (28). Moreover, several recent studies have shown strong evidence of post-admixture selection associated with Khoe and San gene flow in the extended HLA region (27,29). In all the GWAS studies reported to date, the strongest association with RA has been observed in the HLA region, which makes the study of RA in Southern Africans especially relevant due to the possibility that recent post-admixture selection in this genomic region could have impacted the association signals.
This study identified a well-established lead variant near the HLA-DRB1 gene, suggesting an overarching similarity in the genetic architecture of RA across different populations, especially with respect to the HLA region. Imputation using HLA-specific panel, conditional analysis and meta-analysis further supported the presence of at least one additional association peak within the HLA region in African ancestry populations. Given the intrinsic complexities of the region, the findings motivate further studies using more nuanced approaches, such as long-read sequencing and direct HLA-typing for a comprehensive characterization of RA associations in this region.
The GWAS in SEBSSA identified several suggestive associations including a strong signal in a long non-coding RNA on chromosome 4. An interesting suggestive signal, rs60238262 near CTD-2277 K2.1 has been predicted to be an eQTL and has been shown to be expressed in cultured fibroblast cells (https://gtexportal.org/home/). Similarly, the suggestive signal rs11580729 near the DDR2 gene is also a potential eQTL and the functional impact of these signals in terms of gene expression modulations is plausible.
To enhance the statistical power of our study, we conducted a meta-analysis of our summary statistics with that from the CLEAR study based on an AA cohort. Among the signals detected at a suggestive threshold in the meta-analysis, NEDD4 is notable. Studies in murine models of keloid have shown that NEDD4 is critical for NF-κB-dependent activation of several key inflammatory cytokines, such as IL-6 (30). Although this study was aimed at understanding a different variant associated with keloid, these cytokines are common to both diseases and F759 knock-in mice also develop spontaneous arthritis later in life.
Comparisons between the effect size estimates and the EAFs between two cohorts showed similar EAF for both the HLA and non-HLA-associated SNPs, but some differences in effect size estimates, especially for the non-HLA SNPs. In addition to genetics, factors such as lifestyle, behavioral and environmental (for example pathogen triggers) differences in two distinct socio-demographic contexts could have contributed to some of the cohort-specific associations. The results, therefore, indicate that GWAS based on AA populations do not adequately represent the genetics of RA in African populations and underlines the need for a better representation of continental African populations in such studies.
Since there is no standard minimum cut-off for replicating association signals, given our modest sample size, we decided on a P-value cut-off of 5 × 10−3 for exact replication, based on an approach similar to Kuchenbaeckar et al. (31). Despite limited statistical power (Supplementary Material, Fig. S10), we were able to replicate two previously RA-associated signals in the MMEL1 and ANKRD55 genes, showing the replication of established non-HLA signals in the SEBSSA cohort. With increased sample size and detection power in the meta-analysis, we were able to replicate three additional non-HLA associations in African ancestry populations.
Moreover, regional replication based on +/− 100 kb windows of the previously reported RA associations identified signals near genes such as PADI4 and CD28. PADI genes code for peptidylarginine deiminases that catalyze post-translational modification of arginine to citrulline. This process of citrullination plays an important role in inducing ACPA in RA (32).
Although our P-value thresholds might be sub-optimal for providing robust evidence for replication, given our small sample size we expected this approach to enable us to highlight SNPs and genomic regions of probable interest. Future studies and meta-analysis based on larger sample sizes are necessary to replicate some of the other non-HLA associations and reveal the extent of commonality in genetic architecture of RA in populations across continents.
Apart from the inability to detect modest-effect RA-associated SNPs, one of the limitations of the study was that the controls were population-based and not specifically screened for the absence of RA before commencement of the study. Given the relatively low prevalence of RA in the general population, it is unlikely that this might have a major impact on the analysis. Although we employed sex as a covariate in association testing, we recognize that the sex distribution of patients and controls was suboptimal. Since only 14% of our cases were males, we were unable to explore this further with a sex-stratified analysis.
Despite these limitations, the study replicated both HLA and non-HLA associations in a South African cohort and detected several possible novel associations. Future research should aim to develop larger studies to generate an enhanced understanding of the genetic basis of RA across sub-Saharan Africa.
Materials and methods
Patients and controls
Unrelated consenting adult SEBSSA seropositive RA patients, at least 18 years of age at symptom onset were recruited from a tertiary Rheumatology Service in Soweto, Johannesburg. All patients met the 2010 ACR/EULAR classification criteria for RA and were seropositive for ACPA. SEBSSA ethnicity was self-reported and participants confirmed that all four grandparents were of SEBSSA ancestry. The study was approved by the Human Research Ethics Committee (Medical), University of the Witwatersrand (M1706109).
Patient demographics including age, sex, smoking history and serological data (ACPA and RF) were retrieved from the Measurement of Efficacy of Treatment in the Era of Outcome in Rheumatology database (http://www.meteorfoundation.com/) and/or clinical records. Control samples, matched for age and ethnicity, were drawn from the Africa Wits-INDEPTH Partnership for Genomic Studies (AWI-Gen) project, in which participants were recruited from the general population of Soweto and Dikgale, in South Africa (33). The AWI-Gen study was approved by the Human Research Ethics Committee (Medical), University of the Witwatersrand (M121029).
Genotyping and imputation
DNA was extracted using the salting out method (34). Genotyping was performed at Illumina® FastTrack™ Microarray services (Illumina, San Diego, USA) using the H3Africa Consortium SNP genotyping array, which was specifically designed to be enriched for common African variants and included ~ 2.3 million SNPs. Genotype data quality control (QC) was performed using the H3ABioNet/H3Agwas pipeline (https://github.com/h3abionet/h3agwas). SNPs showing missingness greater than 0.05, extreme deviation from Hardy Weinberg equilibrium (P-value<0.0001), and minor allele frequency (MAF) lower than 0.01 were removed. Individuals level QC included exclusion of individuals with missingness greater than 0.05, extreme heterozygosity and sex discrepancy. Related samples with PIHAT>0.18 and PCA outliers were also excluded. Additional genotypes were imputed using the African Genome Resources reference panel (EAGLE2 + PBWT pipeline) at the Sanger Imputation Server (https://imputation.sanger.ac.uk/). Post imputation QC involved restricting the dataset to common (SNPs with MAF ≥ 0.05) and well-imputed SNPs (Info score ≥ 0.8). After QC, 7 908 047 autosomal SNPs were retained for final analysis. In addition, population stratification as a potential confounder between the cases and controls was assessed using PCA using the smartPCA program implemented in EIGENSTRAT (35). PCA with representative African ancestry populations from the 1000 Genomes Project study (24) was generated using the same approach.
Association testing
Genome-wide association testing was conducted using logistic regression, assuming an additive model, in PLINK v1.9 (36,37). A decrease in magnitude of the first 10 PCs was assessed using a scree plot. Based on this, the first three PCs were included as covariates along with sex, age and smoking status. The potential confounding of statistical analysis by population structure was assessed by generating Quantile–Quantile (Q–Q) plots. The data were visualized using the R statistical software (v3.5.1) packages like ‘qqman’ for Manhattan plots (38) and ‘hudson’ for Miami plots (https://github.com/anastasia-lucas/hudson). The level of population structure was also assessed using the genomic inflation factor or lambda (λ), which compares the median distribution of the test for association over that of the null distribution (39,40).
Conditional analysis
To identify possible independent signals within the HLA region, we performed a multi-SNP-based conditional and joint association analysis (COJO) using the program, GCTA (22). This analysis starts with the top SNP (lowest P-value) in a summary statistics file and re-estimates the P-values of the remaining SNPs by conditioning on this. The program then iterates this step and adds top SNPs until no further SNP can be added. We used the SEBSSA dataset as the LD panel for COJO.
Imputation of HLA region with the multi-ethnic reference panel
To improve the imputation of the HLA region, we independently imputed the subset of the SEBSSA genotype dataset corresponding to this region using the multi-ethnic HLA reference panel (based on 36 586 haplotypes from 5 global populations) (23) at the Michigan Imputation Server (https://imputationserver.sph.umich.edu/index.html).
Functional annotation
The likely biological function of genome-wide significantly associated SNPs (P ≤ 5 × 10−8) was assessed using the publicly available web-based tool Ensembl Variant Effect Predictor (http://grch37.ensembl.org/Tools/VEP) (41). The assessment of independence of signals based on 1000 Genomes Project African population LD was performed using FUMA (https://fuma.ctglab.nl/) (42) and genomic regions were visualized using Locuszoom (43). eQTL mapping and tissue expression profiling were also performed using FUMA and GTEX (https://gtexportal.org/home/).
Meta-analysis
Meta-analysis of summary statistics of the current study and a GWAS of AAs with RA (916 RA cases and 1370 controls) as part of the Consortium for the Longitudinal Evaluation of AA with RA (CLEAR) registry (13) was performed using METASOFT (44). For this analysis, only the 6 210 735 SNPs shared between the two studies were considered. We checked if both the alleles were the same in the two studies and excluded SNPs that were a mismatch. Next, we checked if the effect allele in SEBSSA matched the effect allele in the CLEAR study. In case of mismatch, a standardization step (i.e. swapping of alleles accompanied by flipping the direction of the odds ratio for the CLEAR study, to ensure that both studies called association on the same allele) preceded the meta-analysis.
Allele frequency and effect size comparisons
Allele frequencies and effect sizes of the SNPs in the CLEAR study summary statistics were compared with the SEBSSA cohort. Allele frequencies for the SEBSSA cohort were estimated using PLINK v1.9 (36,37). Effect size estimates were based on the association analysis described above.
Replication of known signals
To assess replication of previous RA associations, the GWAS Catalog (25) was mined for studies using ‘Rheumatoid Arthritis’ as the search term. The chromosomal coordinates of the signals were then converted to build 37. SNPs identified as signals in the CLEAR study (13) were excluded from this list, as were non-autosomal SNPs and SNPs without chromosomal coordinates. The final SNP set used for replication included 316 SNPs, of which 46 were from the HLA region.
Threshold for detecting replication
To determine the P-value threshold for detecting replication of known associations, we performed a score test similar to Kuchenbaeckar et al. (31). For the replication of individual signals (exact replication), we randomly sampled 1000 10 kb windows and assessed the lowest P-value in each. We observed P-values lower than 0.005 in <5% of the cases. Based on this estimation we decided on a P-value threshold. A similar analysis with 50 kb windows yielded the same cut-off of 0.001 as observed in the Kuchebaekar et al. 2019 study, suggesting these cut-offs might be portable across studies. For regional replication, we employed 200 kb windows (100 kb on each side of the lead SNP). Score test identified the P-value cut-off of 2.7 × 10−4 to correspond to this window size.
Power calculation
Power for detecting association across the allele frequency range of 0.01– 0.50 and odds ratio range of 1.2–2, for the sample sizes used in the SEBSSA cohort was estimated using the GWAS Power Calculator tool (https://csg.sph.umich.edu/abecasis/cats/gas_power_calculator/). The power estimates are based on the global prevalence rate of 0.46% estimated by Almutairi et al. (45).
Supplementary Material
Acknowledgements
We would like to thank the Chris Hani Baragwanath Academic Hospital Rheumatoid Arthritis Clinic and AWI-Gen participants. M.R. is a South African Research Chair in Genomics and Bioinformatics of African populations hosted by the University of the Witwatersrand, funded by the Department of Science and Technology and administered by the National Research Foundation.
Contributor Information
Evans M Mathebula, Division of Human Genetics, School of Pathology and National Health Laboratory Services, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, 2000, South Africa; Sydney Brenner Institute for Molecular Bioscience, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, 2193, South Africa.
Dhriti Sengupta, Sydney Brenner Institute for Molecular Bioscience, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, 2193, South Africa.
Nimmisha Govind, Sydney Brenner Institute for Molecular Bioscience, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, 2193, South Africa; Division of Rheumatology, University of the Witwatersrand, Johannesburg, 1864, South Africa.
Vincent A Laufer, Division of Clinical Immunology and Rheumatology, University of Alabama at Birmingham (UAB), Birmingham, AL 35294, USA; University of Alabama at Birmingham Medical Scientist Training Program (UAB MSTP), Birmingham, AL 35294, USA; Department of Medicine, University of Alabama at Birmingham, Birmingham, AL 35294, USA.
S Louis Bridges Jr, Department of Medicine, Hospital for Special Surgery, New York, NY, USA and Division of Rheumatology, Weill Cornell Medicine, New York, NY 10021, USA.
Mohammed Tikly, Division of Rheumatology, University of the Witwatersrand, Johannesburg, 1864, South Africa.
Michèle Ramsay, Division of Human Genetics, School of Pathology and National Health Laboratory Services, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, 2000, South Africa; Sydney Brenner Institute for Molecular Bioscience, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, 2193, South Africa.
Ananyo Choudhury, Sydney Brenner Institute for Molecular Bioscience, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, 2193, South Africa.
Conflict of Interest statement
The authors declare no conflict of interest.
Funding
National Research Foundation, South Africa and Connective Tissues Fund, University of the Witwatersrand, South Africa. The AWI-Gen study provided the control participant genotyping data for this study and was funded by the National Institutes of Health (NIH) (U54HG006938).
References
- 1. Silman, A.J. and Pearson, J.E. (2002) Epidemiology and genetics of rheumatoid arthritis. Arthritis Res. Ther., 4, 1–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Mody, G.M., Shaw, J. and Ramchurren, A. (1988) Rheumatoid arthritis impact survey. S. Afr. Med. J., 74, 409–410. [PubMed] [Google Scholar]
- 3. Kalla, A.A. and Tikly, M. (2003) Rheumatoid arthritis in the developing world. Best Pract. Res. Clin. Rheumatol., 17, 863–875. [DOI] [PubMed] [Google Scholar]
- 4. Firestein, G.S. (2003) Evolving concepts of rheumatoid arthritis. Nature, 423, 356–361. [DOI] [PubMed] [Google Scholar]
- 5. Klareskog, L., Stolt, P., Lundberg, K., Källberg, H., Bengtsson, C., Grunewald, J., Rönnelid, J., Harris, H.E., Ulfgren, A.-K., Rantapää-Dahlqvist, S. et al. (2006) A new model for an etiology of rheumatoid arthritis: smoking may trigger HLA-DR (shared epitope)-restricted immune reactions to autoantigens modified by citrullination. Arthritis Rheum., 54, 38–46. [DOI] [PubMed] [Google Scholar]
- 6. van der Woude, D., Houwing-Duistermaat, J.J., Toes, R.E.M., Huizinga, T.W.J., Thomson, W., Worthington, J., van der Helm-van Mil, A.H.M. and de Vries, R.R.P. (2009) Quantitative heritability of anti-citrullinated protein antibody-positive and anti-citrullinated protein antibody-negative rheumatoid arthritis. Arthritis Rheum., 60, 916–923. [DOI] [PubMed] [Google Scholar]
- 7. Singwe-Ngandeu, M., Meli, J., Ntsiba, H., Nouédoui, C., Yollo, A.V., Sida, M.B. and Muna, W.F. (2007) Rheumatic diseases in patients attending a clinic at a referral hospital in Yaounde, Cameroon. East Afr. Med. J., 84, 404–409. [DOI] [PubMed] [Google Scholar]
- 8. Martell, R.W., Stein, M., Davis, P., West, G., Ernmsnuel, J. and du Toit, E.D. (1990) (1990) The association between HLA and rheumatoid artlmtis in Zimbabwean Blacks. The association between HLA and rheumatoid artlmtis in Zimbabwean Blacks. Tissue Antigens, 36, 125–126. [DOI] [PubMed] [Google Scholar]
- 9. Gregersen, P.K., Silver, J. and Winchester, R.J. (1987) The shared epitope hypothesis. An approach to understanding the molecular genetics of susceptibility to rheumatoid arthritis. Arthritis Rheum., 30, 1205–1213. [DOI] [PubMed] [Google Scholar]
- 10. Govind, N., Reynolds, R.J., Hodkinson, B., Ickinger, C., Ramsay, M., Bridges, S.L., Jr. and Tikly, M. (2019) HLA-DRB1 Amino Acid Positions and Residues Associated with Antibody-positive Rheumatoid Arthritis in Black South Africans. J. Rheumatol., 46, 138–144. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Reynolds, R.J., Ahmed, A.F., Danila, M.I., Hughes, L.B., Consortium for the Longitudinal Evaluation of African Americans with Early Rheumatoid Arthritis Investigators, Gregersen, P.K, Raychaudhuri, S., Plenge, R.M. and Bridges, S.L., Jr. (2014) HLA-DRB1-associated rheumatoid arthritis risk at multiple levels in African Americans: hierarchical classification systems, amino acid positions, and residues. Arthritis Rheumatol., 66, 3274–3282. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Govind, N., Choudhury, A., Hodkinson, B., Ickinger, C., Frost, J., Lee, A., Gregersen, P.K., Reynolds, R.J., Bridges, S.L., Jr., Hazelhurst, S. et al. (2014) Immunochip identifies novel, and replicates known, genetic risk loci for rheumatoid arthritis in black South Africans. Mol. Med., 20, 341–349. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Laufer, V.A., Tiwari, H.K., Reynolds, R.J., Danila, M.I., Wang, J., Edberg, J.C., Kimberly, R.P., Kottyan, L.C., Harley, J.B., Mikuls, T.R. et al. (2019) Genetic influences on susceptibility to rheumatoid arthritis in African-Americans. Hum. Mol. Genet., 28, 858–874. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Tikly, M., Govind, N., Frost, J. and Ramsay, M. (2010) The PTPN22 R620W polymorphism is not associated with systemic rheumatic diseases in South Africans. Rheumatology, 49, 820–821. [DOI] [PubMed] [Google Scholar]
- 15. International HapMap Consortium, Frazer, K.A., Ballinger, D.G., Cox, D.R., Hinds, D.A., Stuve, L.L., Gibbs, R.A., Belmont, J.W., Boudreau, A., Hardenbol, P. et al. (2007) A second generation human haplotype map of over 3.1 million SNPs. Nature, 449, 851–861. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Plenge, R.M., Seielstad, M., Padyukov, L., Lee, A.T., Remmers, E.F., Ding, B., Liew, A., Khalili, H., Chandrasekaran, A., Davies, L.R.L. et al. (2007) TRAF1-C5 as a risk locus for rheumatoid arthritis – a genomewide study. N. Engl. J. Med., 357, 1199–1209. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. van Oene, M., Wintle, R.F., Liu, X., Yazdanpanah, M., Gu, X., Newman, B., Kwan, A., Johnson, B., Owen, J., Greer, W. et al. (2005) Association of the lymphoid tyrosine phosphatase R620W variant with rheumatoid arthritis, but not Crohn’s disease, in Canadian populations. Arthritis Rheum., 52, 1993–1998. [DOI] [PubMed] [Google Scholar]
- 18. Yang, X.-K., Liu, J., Liu, J., Liang, Y., Xu, W.-D., Leng, R.-X., Pan, H.-F. and Ye, D.-Q. (2015) Associations between PADI4 gene polymorphisms and rheumatoid arthritis: an updated meta-analysis. Arch. Med. Res., 46, 317–325. [DOI] [PubMed] [Google Scholar]
- 19. Massarenti, L., Enevold, C., Damgaard, D., Ødum, N., Garred, P., Frisch, M., Shelef, M.A., Jacobsen, S. and Nielsen, C.H. (2021) PADI4 polymorphisms confer risk of anti-CCP-positive rheumatoid arthritis in synergy with HLA-DRB1*04 and smoking. Front. Immunol., 12, 707690. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Gerosa, M., De Angelis, V., Riboldi, P. and Meroni, P.L. (2008) Rheumatoid arthritis: a female challenge. Womens Health, 4, 195–201. [DOI] [PubMed] [Google Scholar]
- 21. Reddy, P., Zuma, K., Shisana, O., Kim, J. and Sewpaul, R. (2015) Prevalence of tobacco use among adults in South Africa: results from the first South African National Health and Nutrition Examination Survey. S. Afr. Med. J., 105, 648–655. [DOI] [PubMed] [Google Scholar]
- 22. Yang, J., Lee, S.H., Goddard, M.E. and Visscher, P.M. (2011) GCTA: a tool for genome-wide complex trait analysis. Am. J. Hum. Genet., 88, 76–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Luo, Y., Kanai, M., Choi, W., Li, X., Sakaue, S., Yamamoto, K., Ogawa, K., Gutierrez-Arcelus, M., Gregersen, P.K., Stuart, P.E. et al. (2021) A high-resolution HLA reference panel capturing global population diversity enables multi-ancestry fine-mapping in HIV host response. Nat. Genet., 53, 1504–1516. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. 1000 Genomes Project Consortium, Auton, A., Brooks, L.D., Durbin, R.M., Garrison, E.P., Kang, H.M., Korbel, J.O., Marchini, J.L., McCarthy, S., McVean, G.A. et al. (2015) A global reference for human genetic variation. Nature, 526, 68–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Buniello, A., MacArthur, J.A.L., Cerezo, M., Harris, L.W., Hayhurst, J., Malangone, C., McMahon, A., Morales, J., Mountjoy, E., Sollis, E. et al. (2019) The NHGRI-EBI GWAS Catalog of published genome-wide association studies, targeted arrays and summary statistics 2019. Nucleic Acids Res., 47, D1005–D1012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Ha, E., Bae, S.-C. and Kim, K. (2021) Large-scale meta-analysis across East Asian and European populations updated genetic architecture and variant-driven biology of rheumatoid arthritis, identifying 11 novel susceptibility loci. Ann. Rheum. Dis., 80, 558–565. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Choudhury, A., Aron, S., Botigué, L.R., Sengupta, D., Botha, G., Bensellak, T., Wells, G., Kumuthini, J., Shriner, D., Fakim, Y.J. et al. (2020) High-depth African genomes inform human migration and health. Nature, 586, 741–748. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Sengupta, D., Choudhury, A., Fortes-Lima, C., Aron, S., Whitelaw, G., Bostoen, K., Gunnink, H., Chousou-Polydouri, N., Delius, P., Tollman, S. et al. (2021) Genetic substructure and complex demographic history of South African Bantu speakers. Nat. Commun., 12, 2080. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Patin, E., Lopez, M., Grollemund, R., Verdu, P., Harmant, C., Quach, H., Laval, G., Perry, G.H., Barreiro, L.B., Froment, A. et al. (2017) Dispersals and genetic adaptation of Bantu-speaking populations in Africa and North America. Science, 356, 543–546. [DOI] [PubMed] [Google Scholar]
- 30. Fujita, M., Yamamoto, Y., Jiang, J.-J., Atsumi, T., Tanaka, Y., Ohki, T., Murao, N., Funayama, E., Hayashi, T., Osawa, M. et al. (2019) NEDD4 is involved in inflammation development during keloid formation. J. Invest. Dermatol., 139, 333–341. [DOI] [PubMed] [Google Scholar]
- 31. Kuchenbaecker, K., Telkar, N., Reiker, T., Walters, R.G., Lin, K., Eriksson, A., Gurdasani, D., Gilly, A., Southam, L., Tsafantakis, E. et al. (2019) The transferability of lipid loci across African, Asian and European cohorts. Nat. Commun., 10, 4330. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Darrah, E. and Andrade, F. (2018) Rheumatoid arthritis and citrullination. Curr. Opin. Rheumatol., 30, 72–78. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Ramsay, M., Crowther, N., Tambo, E., Agongo, G., Baloyi, V., Dikotope, S., Gómez-Olivé, X., Jaff, N., Sorgho, H., Wagner, R. et al. (2016) H3Africa AWI-Gen Collaborative Centre: a resource to study the interplay between genomic and environmental risk factors for cardiometabolic diseases in four sub-Saharan African countries. Global Health Epidemiol Genom, 1, e20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Miller, S.A., Dykes, D.D. and Polesky, H.F. (1988) A simple salting out procedure for extracting DNA from human nucleated cells. Nucleic Acids Res., 16, 1215. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Patterson, N., Price, A.L. and Reich, D. (2006) Population structure and eigenanalysis. PLoS Genet., 2, e190. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Chang, C.C., Chow, C.C., Tellier, L.C., Vattikuti, S., Purcell, S.M. and Lee, J.J. (2015) Second-generation PLINK: rising to the challenge of larger and richer datasets. Gigascience, 4, 7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Purcell, S., Neale, B., Todd-Brown, K., Thomas, L., Ferreira, M.A.R., Bender, D., Maller, J., Sklar, P., de Bakker, P.I.W., Daly, M.J. et al. (2007) PLINK: a tool set for whole-genome association and population-based linkage analyses. Am. J. Hum. Genet., 81, 559–575. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Turner, S.D. (2018) qqman: an R package for visualizing GWAS results using Q-Q and manhattan plots. Journal of Open Source Software, 3(25), 731. [Google Scholar]
- 39. Devlin, B. and Roeder, K. (1999) Genomic control for association studies. Biometrics, 55, 997–1004. [DOI] [PubMed] [Google Scholar]
- 40. Hinrichs, A.L., Larkin, E.K. and Suarez, B.K. (2009) Population stratification and patterns of linkage disequilibrium. Genet. Epidemiol., 33(Suppl. 1), S88–S92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. McLaren, W., Gil, L., Hunt, S.E., Riat, H.S., Ritchie, G.R.S., Thormann, A., Flicek, P. and Cunningham, F. (2016) The Ensembl Variant Effect Predictor. Genome Biol., 17, 122. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Watanabe, K., Taskesen, E., van Bochoven, A. and Posthuma, D. (2017) Functional mapping and annotation of genetic associations with FUMA. Nat. Commun., 8, 1826. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Pruim, R.J., Welch, R.P., Sanna, S., Teslovich, T.M., Chines, P.S., Gliedt, T.P., Boehnke, M., Abecasis, G.R. and Willer, C.J. (2010) LocusZoom: regional visualization of genome-wide association scan results. Bioinformatics, 26, 2336–2337. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Han, B. and Eskin, E. (2011) Random-effects model aimed at discovering associations in meta-analysis of genome-wide association studies. Am. J. Hum. Genet., 88, 586–598. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Almutairi, K., Nossent, J., Preen, D., Keen, H. and Inderjeeth, C. (2021) The global prevalence of rheumatoid arthritis: a meta-analysis based on a systematic review. Rheumatol. Int., 41, 863–877. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.