HLA-focused type 1 diabetes genetic risk prediction in populations of diverse ancestry

HLA-focused type 1 diabetes genetic risk prediction in populations of diverse ancestry | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search HLA-focused type 1 diabetes genetic risk prediction in populations of diverse ancestry View ORCID Profile Dominika A. Michalek , View ORCID Profile Courtney Tern , View ORCID Profile Catherine C. Robertson , View ORCID Profile Wei-Min Chen , View ORCID Profile Suna Onengut-Gumuscu , View ORCID Profile Stephen S. Rich doi: https://doi.org/10.1101/2025.08.07.25333167 Dominika A. Michalek 1 Department of Genome Sciences, University of Virginia , Charlottesville, VA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Dominika A. Michalek Courtney Tern 1 Department of Genome Sciences, University of Virginia , Charlottesville, VA, USA 2 Channing Division of Network Medicine, Brigham and Women’s Hospital , Boston, MA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Courtney Tern Catherine C. Robertson 1 Department of Genome Sciences, University of Virginia , Charlottesville, VA, USA 3 Department of Computational Medicine and Bioinformatics, University of Michigan , Ann Arbor, MI, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Catherine C. Robertson Wei-Min Chen 1 Department of Genome Sciences, University of Virginia , Charlottesville, VA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Wei-Min Chen Suna Onengut-Gumuscu 1 Department of Genome Sciences, University of Virginia , Charlottesville, VA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Suna Onengut-Gumuscu Stephen S. Rich 1 Department of Genome Sciences, University of Virginia , Charlottesville, VA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Stephen S. Rich For correspondence: ssr4n{at}virginia.edu Abstract Full Text Info/History Metrics Supplementary material Data/Code Preview PDF Abstract Aims/hypothesis Type 1 diabetes is characterized by the destruction of pancreatic beta cells. Genetic factors account for ~50% of the total risk, with variants in the Human Leukocyte Antigen (HLA) region contributing to half of this genetic risk, with research historically focused on populations of European ancestry. We developed HLA-focused type 1 diabetes genetic risk scores (T1D GRS HLA ) utilizing single nucleotide polymorphisms (SNPs) or HLA alleles from four ancestry groups (Admixed African (AFR; T1D GRS HLA-AFR ), Admixed American (AMR; T1D GRS HLA-AMR ), European (EUR; T1D GRS HLA-EUR ), Finnish (FIN; T1D GRS HLA-FIN ) and across ancestry (ALL; T1D GRS HLA-ALL ). We assessed the performance of genetic risk scores in each population to determine transferability of constructed scores. Methods A total of 41,689 samples and 13,695 SNPs in the HLA region were genotyped, with HLA alleles imputed using the HLA TAPAS multi-ethnic reference panel. Conditionally independent SNPs and HLA alleles associated with type 1 diabetes were identified in each population group to construct T1D GRS HLA models. Generated T1D GRS HLA models were used to predict HLA-focused type 1 diabetes genetic risk across four ancestry groups. Performance of each T1D GRS HLA model was assessed using Receiver Operating Characteristic (ROC) Area Under the Curve (AUC) and compared statistically. Results Each T1D GRS HLA model included a different number of conditionally independent HLA region SNPs (AFR, n = 5; AMR, n = 3; EUR, n = 38; FIN, n = 6; ALL, n = 36) and HLA alleles (AFR, n = 6; AMR, n = 5; EUR, n = 40; FIN, n = 8; ALL, n = 41). The ROC AUC values of T1D GRS HLA from SNPs or HLA alleles were similar and ranged from 0.73 (T1D GRS HLA-Allele-AMR applied to FIN) to 0.88 (T1D GRS HLA-Allele-EUR to EUR). The ROC AUC using the combined set of conditionally independent SNPs (T1D GRS HLA-SNP-ALL ) or HLA alleles (T1D GRS HLA-Allele-ALL ) performed uniformly well across all ancestry groups, ranging from 0.82 to 0.88 for SNPs, and 0.80 to 0.87 for HLA alleles. Conclusions/interpretation T1D GRS HLA models derived from SNPs performed equivalently to those derived from HLA alleles across ancestries. In addition, T1D GRS HLA-SNP-ALL and GRS HLA-Allele-ALL models had consistently high ROC AUC values when applied across ancestry groups. Larger studies in more diverse populations are needed to better assess the transferability of T1D GRS HLA across ancestries. What is already known about this subject? The HLA region harbors 50% of the genetic risk in type 1 diabetes. The majority of the type 1 diabetes genetic risk scores (T1D GRS) used in the field were developed based on European ancestry data. What is the key question? Should genetic risk scores for type 1 diabetes be derived separately for each population, or can a score derived from multiple ancestries effectively capture the complexity of the HLA region and predict the risk of type 1 diabetes across diverse groups? What are the new findings? T1D GRS HLA derived from SNPs performed equivalently to T1D GRS HLA derived from HLA alleles across ancestries. T1D GRS HLA derived from combined ancestry (ALL) data performed equivalent to, or better than, ancestry-specific T1D GRS HLA models developed for each ancestry group. How might this impact on clinical practice in the foreseeable future? The implementation of a single T1D GRS HLA-SNP-ALL constructed from multiple ancestral populations can be a useful tool for population screening aiming to identify individuals at risk for type 1 diabetes for intervention and prevention. Introduction Genetic variation in HLA class I and HLA class II genes are integral to modulating immune responses [ 1 , 2 ]. Numerous studies have identified a wide array of associations involving HLA alleles, with links to infectious diseases [ 3 , 4 ], outcomes of organ transplantation [ 5 ], and risk of autoimmune diseases [ 6 – 8 ]. Type 1 diabetes results from the autoimmune destruction of the insulin-producing beta cells in the pancreas [ 9 ], requiring exogenous insulin for survival. Twin and family studies demonstrate that genetic factors contribute to ~50% of the risk of type 1 diabetes [ 10 , 11 ], with variation in HLA class I (HLA-A, HLA-B, HLA-C) and HLA class II (HLA-DRB1, HLA-DQA1, HLA-DQB1) accounting for 30%-50% of genetic risk [ 12 ]. Epidemiologic studies suggest that the prevalence of type 1 diabetes in children is ~4/1000 in populations of European (EUR) ancestry [ 13 ], with the highest prevalence in Scandinavia (e.g., Finland [ 14 ]) as well as Kuwait [ 15 ]. Lower prevalence in children of African (AFR), Admixed American (AMR), South Asian (SAS), and East Asian (EAS) genetic ancestry has been observed; however, the SEARCH study reported an increasing prevalence of type 1 diabetes in non-EUR ancestry populations in the United States [ 16 ]. The inclusion of non-EUR populations in genetic studies has been limited [ 17 , 18 ], particularly regarding type 1 diabetes [ 19 , 20 ], which is relatively rare in children of other genetic ancestries. More recent studies have expanded to include diverse populations, resulting in the identification of novel loci and new risk variants in known loci [ 21 ]; however, the role of HLA genetic variation on risk of type 1 diabetes is high for EUR ancestry [ 22 ], although ancestry-specific HLA alleles and variants have been identified in non-EUR populations [ 21 , 23 – 26 ]. A growing effort to utilize genetics in risk assessment has led to the development of genetic risk scores (GRSs). Inclusion of genetic variants weighted by their impact on risk provides a single number that can be applied for individualized risk assessment. The initial type 1 diabetes genetic risk score (T1D GRS) [ 27 , 28 ] consisted of single nucleotide polymorphisms (SNPs) at both HLA and non-HLA sites in the genome. Many of the SNPs in the HLA region “tagged” known HLA alleles associated with type 1 diabetes risk. Although several forms of the T1D GRS have been proposed, the T1D GRS1 exhibited good performance in classifying individuals at increased risk of type 1 diabetes, distinguishing risk of type 1 diabetes from type 2 diabetes, and detecting those with type 2 diabetes who require insulin for glucose control [ 27 ]. The subsequent T1D GRS2 incorporated SNPs within the HLA region to estimate known heterozygote effects (e.g., the HLA-DR3/DR4 increased risk) [ 28 ]; however, T1D GRS1 and T1D GRS2 are based primarily on data from EUR ancestry. Recent work demonstrated that an African-American T1D GRS (seven SNPs) outperformed the T1D GRS1 in AFR ancestry populations and was equivalent to the T1D GRS1 in EUR ancestry populations [ 23 ]. There is a need to address the transferability of the impact of genetic variation in the HLA region on genetic risk in diverse ancestry groups. In this study, we utilize dense genotyping in the HLA region in samples of EUR, AFR, AMR, and Finnish (FIN) collections to develop ancestry-specific T1D GRS HLA and a combined T1D GRS HLA based upon SNPs and their imputed HLA alleles. Each T1D GRS HLA is applied to all populations to determine performance in each ancestry group and whether a combined (cosmopolitan) or series of ancestry-specific T1D GRS HLA would be appropriate for global population screening. Furthermore, we aim to assess the difference between using HLA SNPs and HLA alleles to predict T1D risk, to determine whether SNPs serve as adequate proxies for HLA alleles. SNP genotyping is currently more affordable, globally more accessible, and simpler to perform than HLA typing, making it a more practical option for large-scale screening. Our findings suggest that the combined T1D GRS HLA using SNPs may enable consistently high-risk prediction across populations. Methods and Materials Participants Unrelated individuals were assembled through the Type 1 Diabetes Genetics Consortium (T1DGC) for use in genetic studies with the custom fine-mapping ImmunoChip genotyping array [ 29 ]. The dataset included 16,198 cases with type 1 diabetes and 25,491 controls, collected mainly in the US and Europe (ESM Table 1 ). Supergroups of genetic ancestry (EUR, AFR, AMR and EAS) were defined based on k-means clustering [ 21 ]. The FIN population was defined within the EUR supergroup using k-means clustering. Principal components (PCs) for clustering were derived from projecting samples onto the 1000 Genomes Project phase 3 Reference Panel using PLINK v1.9 [ 30 ]. The South Asian (SAS) ancestry group was separated from the AMR supergroup using multi-dimensional scaling (MDS) analysis implemented in KING v2.3.2 [ 31 ]. Together, the clustering resulted in six genetic ancestry groups (EUR, AFR, AMR, FIN, SAS, EAS). To control for population structure, within each group, PCs were generated using PLINK v1.9 [ 30 ] by performing principal component analysis (PCA) in unrelated control individuals and projecting the subjects with type 1 diabetes onto controls. In addition, samples from all genetic ancestry groups (ALL) were combined, and PCs were generated by projection of all samples onto 1000 Genomes Reference Panel. For PCA, we excluded regions of high linkage disequilibrium (LD) [ 32 ], pruned for short-range LD (r 2 > 0.2 in 50-kb windows) and removed SNPs with minor allele frequency (MAF) ≤ 0.05. Individuals assigned to SAS and EAS were excluded from the primary analyses due to small sample size. View this table: View inline View popup Download powerpoint Table 1. Performance of HLA-focused T1D GRS HLA-SNP-ALL and GRS HLA-Allele-ALL compared to ancestry-specific T1D GRS HLA in each ancestry. T1D diagnosis – eligibility criteria Diagnosis of T1D is critical in determining precision of genetic risk scores. In this analysis, the T1D cases were obtained from two primary sources, the Type 1 Diabetes Genetics Consortium (T1DGC) and the SEARCH for Diabetes in Youth Study. Exclusion of T1D (control samples) was based upon absence of a diagnosis and insulin use. In the T1DGC, the eligibility criteria were (a) diagnosis before 35 years of age, (b) the use of insulin within 6 months of diagnosis, and (c) continuous use of insulin without stopping for 6 months or more. If any question of diagnosis occurred, a Clinical Committee evaluated available records for decision on inclusion. In the SEARCH study, a case was defined by a subject under 20 years of age with physician-diagnosed diabetes, confirmed by medical records review and/or provider reports, an in-person visit, with follow-up laboratory testing (including C-peptide). Genotyping Genotype data in the HLA region on human chromosome 6 (28Mbp - 34Mbp) were extracted from the ImmunoChip panel, with quality control performed as previously published [ 21 ]. Briefly, DNA samples were genotyped on the Illumina ImmunoChip, raw genotyping files were assembled, and genotype clusters were generated using the Illumina GeneTrain2 algorithm. Variant filters were applied to (1) re-annotate positions by aligning probe sequences to GRCh37 (hg19) and the removal of any variants with <100% match or multiple matches at different positions in the genome; (2) removal of variants with call rates <98%; (3) removal of variants with any discordance between duplicate or monozygotic twin samples, as confirmed by genotype-inferred relationships; (4) removal of variants with Mendelian inconsistencies in >1% of the informative trios or parent–offspring pairs, based on genotype-inferred relationships. X-chromosome heterozygosity and Y-chromosome missingness identified (and used to exclude) participants with apparent sex-chromosome anomalies or inconsistencies with the reported sex using KING version 2.1.3 [ 31 ]. Samples with a genotype call rate⍰<⍰98% were removed. Variants with genotype frequencies deviating from Hardy–Weinberg equilibrium ( P ⍰<⍰5⍰×⍰10 −5 ) in unrelated EUR-ancestry controls were excluded before imputation. HLA Imputation HLA imputation was performed using HLA-TAPAS implemented on the University of Michigan imputation server ( https://imputationserver.sph.umich.edu/index.html ) [ 33 ]. SNPs in the HLA region (28Mbp - 34Mbp) were used to predict classical alleles for HLA class I (HLA- A , - B , and - C ) and HLA class II (HLA- DQA1 , - DQB1 , - DRB1 , - DPA1 , and - DPB1 ) genes with two-field accuracy. For quality control after imputation, any variant with imputation accuracy of r 2 ≤ 0.5 and MAF ≤ 0.005 was removed from further analysis. All coordinates are reported in GRCh37. Association Analysis with Type 1 Diabetes to Select SNPs and HLA Alleles for T1D GRSHLA SNPs and HLA class I and HLA class II alleles were analyzed for association with type 1 diabetes within each ancestry group (AFR, AMR, EUR, FIN) and using the combined dataset (ALL) from these groups. Logistic regression models were implemented in PLINK v.1.9, adjusting for five ancestry-specific principal components and using a minor allele count (MAC) ≥ 20 filter. Conditional analyses were performed on SNPs and HLA alleles to identify statistically independent contributors to type 1 diabetes in each group (AFR, AMR, EUR, FIN and ALL) separately. A list of conditionally independent variants was developed by (1) including the most associated variant in the logistic regression model and (2) progressively incorporating the next most significant variants until reaching the last variant that surpassed the significance threshold. For each ancestry group, statistical significance was determined using a Bonferroni-corrected P-value threshold (alpha = 0.05) correcting for the total number of SNPs (P < 3.5 × 10 −6 ) and HLA alleles (P < 3.6 × 10 −4 ). For the combined dataset (ALL), statistical significance for SNPs was P < 3.4 × 10 −6 and P < 2.8 × 10 −4 for HLA alleles. HLA-focused Type 1 Diabetes Genetic Risk Score (T1D GRSHLA) Within each ancestry group (AFR, AMR, EUR, FIN) and across-ancestry groups (ALL) HLA-focused type 1 diabetes genetic risk scores (T1D GRS HLA ) were developed using two separate genetic markers (1) SNPs (T1D GRS HLA-SNP ) and (2) classical HLA alleles (T1D GRS HLA-Allele ). Within each group (AFR, AMR, EUR, FIN and ALL), to capture genetic contribution of the HLA region to type 1 diabetes, SNPs identified from conditional analysis were selected to create T1D GRS HLA-SNP-AFR , T1D GRS HLA-SNP-AMR , T1D GRS HLA-SNP-EUR , T1D GRS HLA-SNP-FIN and T1D GRS HLA-SNP-ALL and HLA alleles identified from conditional analyses were used to create T1D GRS HLA-Allele-AFR , T1D GRS HLA-Allele-AMR , T1D GRS HLA-Allele-EUR , T1D GRS HLA-Allele-FIN and T1D GRS HLA-Allele-ALL . In total, five T1D GRS HLA models were developed for each type of genetic marker. ESM Tables 2-6 provide list of SNPs included in each T1D GRS HLA-SNP model (including information on effect allele, allele frequency and regression coefficient) and similarly, ESM Tables 7-11 provide the list of HLA alleles included in each T1D GRS HLA-Allele model. The regression coefficients for each SNP (or classical HLA allele) included in the GRS HLA model were used as weights for the individual SNPs (or HLA alleles). A GRS HLA for each individual was calculated by summing the allele counts multiplied by their respective weight (KING v.2.3.2). Transferability of T1D GRSHLA in SNPs and HLA Alleles Each T1D GRS HLA model (AFR, AMR, EUR, FIN and ALL), constructed from SNPs and HLA alleles, was applied to individuals in all ancestry groups (AFR, AMR, EUR, FIN). All participants in each ancestry cohort would have five T1D GRS HLA-SNP values, and five T1D GRS HLA-Allele values. To determine the equivalence of T1D GRS HLA risk prediction across ancestry-specific T1D GRS HLA-SNP and T1D GRS HLA-Allele models, Receiver Operating Characteristic (ROC) Area Under the Curve (AUC) values were computed using pROC R package [ 34 ]. The test for comparing two AUC values used the method of DeLong to determine equivalence of AUC values [ 35 , 36 ]. Results A total of 41,689 samples and 13,695 SNPs genotyped in the HLA region were included in the study, comprising 16,198 individuals with type 1 diabetes (cases) and 25,491 individuals without type 1 diabetes (controls). For each ancestry group, 425 HLA alleles were imputed at 2-field resolution using a multi-ethnic HLA reference panel (HLA-TAPAS). SNPs were projected onto 1000 Genomes Project reference populations, and individuals were assigned to EUR (N = 33,601), AFR (N = 3,877), AMR (N = 1,084), and FIN (N = 2,804) genetic ancestry groups. Individuals assigned to SAS (N = 179) and EAS (N = 144) were excluded from the primary analyses due to small sample sizes, limiting the final sample size to 41,366 participants. Conditionally independent HLA region variants associated with type 1 diabetes risk We tested SNPs and classical HLA alleles within the MHC region to identify variants that have independent effects on type 1 diabetes risk. The most significantly associated SNP with type 1 diabetes, across all ancestries and combined data, was rs9273363 (OR AFR = 5.56, P AFR = 1.04 × 10 −133 , OR AMR = 3.72, P AMR = 1.09 × 10 −38 , OR EUR = 4.81, P EUR = 8.36 × 10 −1464 , OR FIN = 3.64, P FIN = 1.19 × 10 −88 , OR ALL = 4.76, P ALL = 3.15 × 10 −1738 ). When we tested HLA alleles for association, in AFR and AMR ancestry the most significant association with risk was with HLA- DQA1*03:01 (OR AFR = 5.45, P AFR = 9.28 × 10 −116 ; OR AMR = 2.91, P AMR = 2.44 × 10 −21 ); while in EUR and FIN ancestry HLA- DQB1*03:02 was most strongly associated with type 1 diabetes risk (OR EUR = 5.33, P EUR = 5.08 × 10 −1145 ; OR FIN = 3.91, P FIN = 8.76 × 10 −76 ). The most significantly associated HLA allele in the combined ancestry data was HLA- DQB1*03:02 (OR ALL = 5.13, P ALL = 5.28 × 10 −1314 ). In addition to identifying the most significant variant in each population, we conducted a stepwise conditional analysis to identify additional independently associated type 1 diabetes risk variants. The number of conditionally independent SNPs and HLA alleles is provided below in the GRS models and the details, including allele weights and allele frequencies are provided in ESM Tables 2-6 and ESM Tables 7-11, respectively. HLA-focused type 1 diabetes genetic risk scores (T1D GRSHLA) Genetic risk scores for type 1 diabetes derived from the HLA region (T1D GRS HLA ) were calculated using SNPs (T1D GRS HLA-SNP ) and HLA alleles (T1D GRS HLA-Allele ) across four ancestry populations (AFR, AMR, EUR, FIN) and combined data (ALL) using SNPs and imputed HLA alleles that were independently associated with type 1 diabetes. Of the four ancestry groups, EUR had the largest population size and the EUR-derived T1D GRS HLA models contained the most SNPs (n=38) and HLA alleles (n=40), while other GRS HLA models had less than 10 SNPs and HLA alleles with conditional independence of association with type 1 diabetes. The T1D GRS HLA developed in AFR consists of 5 SNPs and 6 HLA alleles, in AMR of 3 SNPs and 5 HLA alleles, and in FIN of 6 SNPs and 8 HLA alleles (ESM Tables 3-6 and ESM Tables 8-11). The model constructed from combined data (ALL) was dominated by those SNPs and HLA alleles from the EUR population, and the T1D GRS HLA models contained 36 SNPs and 41 HLA alleles (ESM Table 2 and ESM Table 7). Each T1D GRS HLA model (AFR, AMR, EUR, FIN and ALL) was applied to all ancestry groups (AFR, AMR, EUR, FIN). We plotted the distribution of each T1D GRS HLA in all ancestry groups separating T1D cases and controls. The distribution of T1D GRS HLA within all populations exhibited significant overlap in those with type 1 diabetes and those without type 1 diabetes, whether using SNPs ( Fig. 1a ) or imputed HLA alleles ( Fig. 1b ). The bimodal distribution observed in the EUR and FIN populations can be explained by the presence of a substantial number of individuals carrying high-risk HLA haplotypes (HLA-DR3 and/or DR4), who exhibit the highest T1D GRS values. ESM Figure 1 . shows the distribution of T1D GRS HLA-Allele-ALL , applied across all ancestry groups (AFR, AMR, EUR, FIN) and stratified by cases carrying high-risk HLA haplotypes (HLA-DR3 and/or DR4) and control subjects. Download figure Open in new tab Download figure Open in new tab Fig 1. Raincloud plots of HLA-focused type 1 diabetes genetic risk scores (T1D GRS HLA ) using combined ancestry data (ALL) SNPs ( Fig. 1a ) or HLA alleles ( Fig. 1b ) in four ancestry groups. The x-axis represents the HLA-focused T1D GRS HLA by case (T1D - green ( Fig. 1a ), blue ( Fig. 1b )) and control (no T1D - yellow ( Fig. 1a ), pale pink ( Fig. 1b )) status in each ancestry group. The dots below the box plot represent the individual scores, while the distribution is plotted above the box plot. The box plot (median, interquartile range, and range) is shown between the upper and lower distributions. Prediction and transferability of type 1 diabetes using ROC curve analysis Prediction of type 1 diabetes risk (defined by ROC AUC) using SNPs ( Fig. 2a , ESM Table 12) was uniformly high, ranging from 0.74 (T1D GRS HLA-SNP-FIN applied to AMR) to 0.88 (T1D GRS HLA-SNP-ALL applied to EUR). Similarly, the ROC AUC using HLA alleles ( Fig. 2b , ESM Table 13) ranged from 0.73 (T1D GRS HLA-Allele-AMR applied to FIN) to 0.88 (T1D GRS HLA-Allele-EUR to EUR). The T1D GRS HLA model based upon combined data (T1D GRS HLA-SNP-ALL and T1D GRS HLA-Allele-ALL ) performed equivalently to the best individual ancestry-derived models across groups, whether using SNPs or HLA alleles. There were no significant differences in model performance, whether using SNPs or imputed HLA alleles for any comparison. Download figure Open in new tab Download figure Open in new tab Fig 2. Area Under the Curve (AUC) from Receiver-Operator Characteristic (ROC) analyses of HLA-focused T1D GRS HLA models based upon HLA SNPs ( Fig. 2a ) or HLA alleles ( Fig. 2b ) in four ancestry groups. Each HLA-focused T1D GRS HLA model is color-coded based upon the ancestry from which it was derived (AFR - light blue, AMR - olive green, EUR - pink, FIN - light purple, ALL - yellow). Values of AUC per model are presented in the lower right corner for each ancestry group. To evaluate transferability of scores, we compared ancestry-derived T1D GRS HLA scores with T1D GRS HLA-ALL scores within each ancestry group ( Table 1 ). In AFR ancestry group, the performance of T1D GRS HLA-SNP-ALL did not differ significantly from T1D GRS HLA-SNP-AFR (AUC ALL = 0.86 vs. AUC AFR = 0.86, p = 0.11). Similarly, in FIN ancestry group, the performance of T1D GRS HLA-SNP-ALL did not differ significantly from T1D GRS HLA-SNP-FIN (AUC ALL = 0.82 vs. AUC FIN = 0.82, p = 0.79). In contrast, T1D GRS HLA-SNP-ALL performed significantly better in AMR (AUC ALL = 0.82 vs. AUC AMR = 0.78, p = 7.86 × 10 −6 ) and EUR (AUC = 0.88 vs. AUC ALL = 0.87, p EUR = 4.73 × 10 −6 ) ancestry groups. T1D risk prediction in an independent cohort and the effect of non-HLA SNPs In order to address two limitations of cross-ancestry group comparisons, we have (1) included non-HLA region SNPs [ 21 ] to T1D GRS HLA-SNP-ALL model and re-calculated the T1D GRS in each ancestry and (2) conducted a validation study in a population of diverse ancestry. Incorporating non-HLA SNPs in the T1D GRS score improved prediction in all groups. In the AFR group, the AUC increased from 0.86 to 0.88, and in the AMR and FIN groups, the AUC increased from 0.82 to 0.85 and from 0.82 to 0.84, respectively. In EUR group, the AUC increased from 0.88 to 0.91. Together, these results suggest that inclusion of non-HLA SNPs increases the predictive accuracy of the T1D GRS. In a larger, genetically diverse validation cohort (510 T1D cases, 6,342 controls; 30% AFR, 18% AMR, 11% EAS, and 41% EUR), using 23 HLA-region SNPs yielded an AUC = 0.806. The inclusion of 67 non-HLA region SNPs resulted in only a slight increase in predictive performance (AUC = 0.810). Thus, the T1D GRS including all SNPs was validated with a high AUC (~0.80) even though non-HLA SNPs did not significantly improve the AUC beyond that achieved by HLA SNPs alone. Discussion In this study, we constructed HLA-focused T1D Genetic Risk Scores, T1D GRS HLA , in different ancestry groups as well as combining data across populations. To develop genetic risk score models, we identified independently associated SNPs, HLA class I alleles, and HLA class II alleles in Admixed African (AFR), Admixed American (AMR), European (EUR), Finnish (FIN) groups and across-ancestry (ALL). We applied each T1D GRS HLA model to four ancestry groups. Our results suggest that T1D GRS HLA derived from combined ancestry (ALL) data performed equivalent to, or better than, ancestry-specific T1D GRS HLA models defined in each ancestry group. Each T1D GRS HLA model included different numbers of conditionally independent HLA region SNPs associated with type 1 diabetes. The most significantly associated SNP with type 1 diabetes, across all ancestries and combined data, was rs9273363. This SNP has been identified as the most strongly associated with T1D risk and tags HLA- DQB1*03:02 in European ancestry populations. As expected, HLA- DQB1*03:02 was most strongly associated with T1D risk in the EUR and FIN groups, while in the AFR and AMR groups, the most strongly associated HLA class II allele was HLA DQA1*03:01 , which aligns with our previous findings from recent type 1 diabetes GWAS [ 24 ]. Given the much larger EUR sample size in the GWAS (and the current HLA-focused study), the analysis of samples from all ancestry groups also identified HLA -DQB1*03:02 as the most strongly associated HLA allele with T1D risk. It is well known that the HLA region has undergone selective pressure [ 37 ], leading to allele frequency changes across populations. Although many of the SNPs and HLA alleles associated with T1D risk with respect to the HLA region are shared across ancestries, the size of the effect on risk can vary dramatically. In this study, the effect of HLA- DQB1*03:02 and the tagging SNP rs9273363 on T1D risk varies by ancestry. For rs9273363, the effect size in AFR (OR = 5.56) is larger than that in AMR (OR = 3.72), EUR (OR = 4.81), and FIN (OR = 3.64). In contrast, the HLA- DQB1*03:02 allele tagged by this SNP differs in AFR (OR = 1.68; the HLA- DQA1*03:01 allele is most strongly associated with T1D risk, with effect OR = 5.45), AMR (not significant in this group), EUR (OR = 5.33), and FIN (OR = 3.91). These results suggest the complexity of the HLA associations with type 1 diabetes and highlight the need to develop scores that include ancestry-informed risk prediction. Applications of genetic variation of disease in the clinic and in population screening have adopted the methodology of genetic risk scores as an approach to reduce complexity into a single metric, similar to any clinical laboratory test (e.g., HDL cholesterol) with comparison to a standard reference range. A major limitation has been the reliance on genetic data from studies of European ancestry [ 17 ]. The field has adapted T1D GRSs [ 27 , 28 ] that have been generated using European-ancestry data. They perform well to distinguish type 1 diabetes from type 2 diabetes, predict progression to insulin deficiency and improve newborn screening. In addition, the T1D GRS1 has been shown to discriminate type 1 diabetes from type 2 diabetes and maturity-onset diabetes of the young (MODY) [ 38 ]. However, there is a critical need to include non-European populations in the development of more accurate genetic risk scores, and to create T1D GRS models that incorporate diverse ancestry populations. Recent applications of T1D GRS [ 23 , 39 – 41 ] and methods development in two consortia (eMERGE and PRIMED) are beginning to address these gaps. In our paper we are recalculating weights for associated SNPs within each ancestry group and across all individuals. This is in comparison to other methods, where weights are driven from European populations. There are increasing efforts to assess the performance of the T1D GRS1 and T1D GRS2, that were developed in EUR-ancestry populations, when applied to other populations. Recent analysis of a 67 SNP T1D GRS in an Indian population discriminated T1D cases from controls (AUC = 0.83) although lower than in EUR populations (AUC = 0.92) [ 42 ]. An updated score (T1D GRS2x) was used in the genetically diverse All of Us Research Program [ 43 ] and was highly predictive (multi-ancestry AUC = 0.860) but lower than in EUR (AUC = 0.895). Utility of multi-ancestry T1D GRSs and development across diverse ancestries show promise, yet increased sample sizes and ancestry groups will provide improvements beyond those of EUR ancestry. This study has several strengths and limitations. Strengths of this study include use of multi-ancestry cohorts, focus on the most important genomic region for T1D risk (the HLA region), imputation to increase number of SNPs and HLA alleles (using HLA-TAPAS), and comparison of ancestry-specific SNP- and HLA allele-based genetic risk scores. However, some limitations include the smaller number of under-represented ancestry-diverse populations (AFR, AMR), and excluding potentially informative populations due to extremely small sample size (EAS, SAS). In addition, not all HLA alleles could be imputed in all populations (e.g., HLA DQB1*02:02 ). Finally, a limitation is that there is an overlap between the training data and the testing data that may affect interpretation of performance. These limitations highlight the need to expand non-EUR cohorts to better assess the prediction of ancestry-specific genetic risk scores for all diseases and, in this case, type 1 diabetes. In summary, our data suggests that T1D GRS HLA derived from SNPs performs equivalently to T1D GRS HLA derived from HLA alleles across ancestries. The T1D GRS HLA model derived from one ancestry is not uniformly predictive in other ancestries. Greater sample sizes in non-EUR ancestry populations are needed to develop more accurate non-EUR T1D GRS HLA . In addition, our results suggest that T1D GRS HLA-ALL constructed from combined ancestry data performed equally well or better than ancestry-specific scores in each ancestry. While we do not have a clear explanation, we believe that the combined ancestry GRS (T1D GRS HLA-ALL ) performed better than the individual ancestry scores, because with greater sample size we have more power to detect T1D associated SNPs. In addition, we are including individuals of diverse ancestry allowing us to tag ancestry specific T1D risk HLA alleles and haplotypes, compared to using European ancestry SNPs only. Funding Research reported in this publication was supported by the National Institutes of Health grants, the Type 1 Diabetes Genetics Consortium (U01DK062418), DP3DK085678, R01DK122586, U01HG011723 (the PRIMED consortium, D-PRISM), the Leona M. and Harry B. Helmsley Charitable Trust (grant #2204-05134), and the Juvenile Diabetes Research Foundation, now known as Breakthrough T1D (grants #1-2001-916 and #9-2011-530.) Data availability Summary statistics are available in the NIH database for Genotype and Phenotype (dbGaP, **link will be provided** ), with accession number **will be provided** , and the Accelerating Medicines Partnership Common Metabolic Diseases (AMP CMD) Knowledge Portal ( https://hugeamp.org/ ). Code used to generate results is available at https://github.com/damichalek/T1DGC_GRS_HLA . Contribution statement SSR and SOG conceptualized, designed the study and contributed to data acquisition. DAM, CT, CCR, WMC conducted statistical analysis. DAM, WMC, SSR and SOG were involved in data interpretation. DAM, SSR and SOG drafted the manuscript. All authors contributed to the editing and critical revision of the manuscript. All authors read and approved the final version to be published. The online version of this article ( https://doi.org/10.1007/xxx ) contains peer reviewed but unedited supplementary material. Acknowledgments The authors express their gratitude to the investigators and groups who collected and provided biological samples or data for this study, as well as to the participants whose contributions made this research possible. Abbreviations AFR Admixed African ALL All Ancestries AMR Admixed American AUC Area Under the Curve EUR European FIN Finnish GRS Genetic Risk Score HLA Human Leukocyte Antigen MAF Minor Allele Frequency MODY Maturity-Onset Diabetes of the Young OR Odds Ratio PCA Principal Component Analysis PC Principal Component ROC Receiver Operating Characteristic SNP Single Nucleotide Polymorphism T1DGC Type 1 Diabetes Genetics Consortium T1D GRS Type 1 Diabetes Genetic Risk Score References 1. ↵ Williams TM ( 2001 ) Human leukocyte antigen gene polymorphism and the histocompatibility laboratory . J Mol Diagn 3 ( 3 ): 98 – 104 . doi: 10.1016/S1525-1578(10)60658-7 OpenUrl CrossRef PubMed Web of Science 2. ↵ Morran MP , Vonberg A , Khadra A , Pietropaolo M ( 2015 ) Immunogenetics of type 1 diabetes mellitus . Molecular Aspects of Medicine 42 : 42 – 60 . doi: 10.1016/j.mam.2014.12.004 OpenUrl CrossRef PubMed 3. ↵ Carrington M , Nelson GW , Martin MP , et al. ( 1999 ) HLA and HIV-1: Heterozygote Advantage and B*35 - Cw*04 Disadvantage . Science 283 ( 5408 ): 1748 – 1752 . doi: 10.1126/science.283.5408.1748 OpenUrl Abstract / FREE Full Text 4. ↵ Thursz MR , Thomas HC , Greenwood BM , Hill AVS ( 1997 ) Heterozygote advantage for HLA class-II type in hepatitis B virus infection . Nat Genet 17 ( 1 ): 11 – 12 . doi: 10.1038/ng0997-11 OpenUrl CrossRef PubMed Web of Science 5. ↵ Anasetti C , Amos D , Beatty PG , et al. ( 1989 ) Effect of HLA compatibility on engraftment of bone marrow transplants in patients with leukemia or lymphoma . N Engl J Med 320 ( 4 ): 197 – 204 . doi: 10.1056/NEJM198901263200401 OpenUrl CrossRef PubMed Web of Science 6. ↵ Singal DP , Blajchman MA ( 1973 ) Histocompatibility (HL-A) antigens, lymphocytotoxic antibodies and tissue antibodies in patients with diabetes mellitus . Diabetes 22 ( 6 ): 429 – 432 . doi: 10.2337/diab.22.6.429 OpenUrl Abstract / FREE Full Text 7. Nerup J , Platz P , Andersen OO , et al. ( 1974 ) HL-A antigens and diabetes mellitus . Lancet 2 ( 7885 ): 864 – 866 . doi: 10.1016/s0140-6736(74)91201-x OpenUrl CrossRef PubMed Web of Science 8. ↵ Cudworth AG , Woodrow JC ( 1975 ) HL-A system and diabetes mellitus . Diabetes 24 ( 4 ): 345 – 349 . doi: 10.2337/diab.24.4.345 OpenUrl Abstract / FREE Full Text 9. ↵ Eisenbarth GS ( 1986 ) Type I diabetes mellitus. A chronic autoimmune disease . N Engl J Med 314 ( 21 ): 1360 – 1368 . doi: 10.1056/NEJM198605223142106 OpenUrl CrossRef PubMed Web of Science 10. ↵ Barnett AH , Eff C , Leslie RD , Pyke DA ( 1981 ) Diabetes in identical twins. A study of 200 pairs . Diabetologia 20 ( 2 ): 87 – 93 . doi: 10.1007/BF00262007 OpenUrl CrossRef PubMed Web of Science 11. ↵ Redondo MJ , Jeffrey J , Fain PR , Eisenbarth GS , Orban T ( 2008 ) Concordance for islet autoimmunity among monozygotic twins . N Engl J Med 359 ( 26 ): 2849 – 2850 . doi: 10.1056/NEJMc0805398 OpenUrl CrossRef PubMed Web of Science 12. ↵ Noble JA , Valdes AM , Cook M , Klitz W , Thomson G , Erlich HA ( 1996 ) The role of HLA class II genes in insulin-dependent diabetes mellitus: molecular analysis of 180 Caucasian, multiplex families . Am J Hum Genet 59 ( 5 ): 1134 – 1148 OpenUrl PubMed Web of Science 13. ↵ Green A , Hede SM , Patterson CC , et al. ( 2021 ) Type 1 diabetes in 2017: global estimates of incident and prevalent cases in children and adults . Diabetologia 64 ( 12 ): 2741 – 2750 . doi: 10.1007/s00125-021-05571-8 OpenUrl CrossRef PubMed 14. ↵ DIAMOND Project Group ( 2006 ) Incidence and trends of childhood Type 1 diabetes worldwide 1990-1999 . Diabet Med 23 ( 8 ): 857 – 866 . doi: 10.1111/j.1464-5491.2006.01925.x OpenUrl CrossRef PubMed Web of Science 15. ↵ Al-Abdulrazzaq D , Qabazard M , Al-Jasser F , et al. ( 2024 ) Early Onset of Type 1 Diabetes in Kuwait: Distinct Clinical, Metabolic, and Immunological Characteristics . Med Princ Pract 33 ( 6 ): 555 – 561 . doi: 10.1159/000540705 OpenUrl CrossRef PubMed 16. ↵ Mayer-Davis EJ , Lawrence JM , Dabelea D , et al. ( 2017 ) Incidence Trends of Type 1 and Type 2 Diabetes among Youths, 2002-2012 . N Engl J Med 376 ( 15 ): 1419 – 1429 . doi: 10.1056/NEJMoa1610187 OpenUrl CrossRef PubMed 17. ↵ Martin AR , Kanai M , Kamatani Y , Okada Y , Neale BM , Daly MJ ( 2019 ) Clinical use of current polygenic risk scores may exacerbate health disparities . Nat Genet 51 ( 4 ): 584 – 591 . doi: 10.1038/s41588-019-0379-x OpenUrl CrossRef PubMed 18. ↵ Fatumo S , Chikowore T , Choudhury A , Ayub M , Martin AR , Kuchenbaecker K ( 2022 ) A roadmap to increase diversity in genomic studies . Nat Med 28 ( 2 ): 243 – 250 . doi: 10.1038/s41591-021-01672-4 OpenUrl CrossRef PubMed 19. ↵ Barrett JC , Clayton DG , Concannon P , et al. ( 2009 ) Genome-wide association study and meta-analysis find that over 40 loci affect risk of type 1 diabetes . Nat Genet 41 ( 6 ): 703 – 707 . doi: 10.1038/ng.381 OpenUrl CrossRef PubMed Web of Science 20. ↵ Kaddis JS , Perry DJ , Vu AN , et al. ( 2022 ) Improving the Prediction of Type 1 Diabetes Across Ancestries . Diabetes Care 45 ( 3 ): e48 – e50 . doi: 10.2337/dc21-1254 OpenUrl CrossRef PubMed 21. ↵ Robertson CC , Inshaw JRJ , Onengut-Gumuscu S , et al. ( 2021 ) Fine-mapping, trans-ancestral and genomic analyses identify causal variants, cells, genes and drug targets for type 1 diabetes . Nat Genet 53 ( 7 ): 962 – 971 . doi: 10.1038/s41588-021-00880-5 OpenUrl CrossRef PubMed 22. ↵ Noble JA ( 2024 ) Fifty years of HLA-associated type 1 diabetes risk: history, current knowledge, and future directions . Front Immunol 15 : 1457213 . doi: 10.3389/fimmu.2024.1457213 OpenUrl CrossRef PubMed 23. ↵ Onengut-Gumuscu S , Chen W-M , Robertson CC , et al. ( 2019 ) Type 1 Diabetes Risk in African-Ancestry Participants and Utility of an Ancestry-Specific Genetic Risk Score . Diabetes Care 42 ( 3 ): 406 – 415 . doi: 10.2337/dc18-1727 OpenUrl Abstract / FREE Full Text 24. ↵ Michalek DA , Tern C , Zhou W , et al. ( 2024 ) A multi-ancestry genome-wide association study in type 1 diabetes . Hum Mol Genet 33 ( 11 ): 958 – 968 . doi: 10.1093/hmg/ddae024 OpenUrl CrossRef PubMed 25. Noble JA , Besançon S , Sidibé AT , et al. ( 2024 ) Complete HLA genotyping of type 1 diabetes patients and controls from Mali reveals both expected and novel disease associations . HLA 103 ( 1 ): e15319 . doi: 10.1111/tan.15319 OpenUrl CrossRef 26. ↵ Gao Y , Chen S , Gu W-Y , et al. ( 2021 ) Genome-wide association study reveals novel loci for adult type 1 diabetes in a 5-year nested case-control study . World J Diabetes 12 ( 12 ): 2073 – 2086 . doi: 10.4239/wjd.v12.i12.2073 OpenUrl CrossRef PubMed 27. ↵ Oram RA , Patel K , Hill A , et al. ( 2016 ) A Type 1 Diabetes Genetic Risk Score Can Aid Discrimination Between Type 1 and Type 2 Diabetes in Young Adults . Diabetes Care 39 ( 3 ): 337 – 344 . doi: 10.2337/dc15-1111 OpenUrl Abstract / FREE Full Text 28. ↵ Sharp SA , Rich SS , Wood AR , et al. ( 2019 ) Development and Standardization of an Improved Type 1 Diabetes Genetic Risk Score for Use in Newborn Screening and Incident Diagnosis . Diabetes Care 42 ( 2 ): 200 – 207 . doi: 10.2337/dc18-1785 OpenUrl Abstract / FREE Full Text 29. ↵ Onengut-Gumuscu S , Chen W-M , Burren O , et al. ( 2015 ) Fine mapping of type 1 diabetes susceptibility loci and evidence for colocalization of causal variants with lymphoid gene enhancers . Nat Genet 47 ( 4 ): 381 – 386 . doi: 10.1038/ng.3245 OpenUrl CrossRef PubMed 30. ↵ Chang CC , Chow CC , Tellier LC , Vattikuti S , Purcell SM , Lee JJ ( 2015 ) Second-generation PLINK: rising to the challenge of larger and richer datasets . Gigascience 4 : 7 . doi: 10.1186/s13742-015-0047-8 OpenUrl CrossRef PubMed 31. ↵ Manichaikul A , Mychaleckyj JC , Rich SS , Daly K , Sale M , Chen W-M ( 2010 ) Robust relationship inference in genome-wide association studies . Bioinformatics 26 ( 22 ): 2867 – 2873 . doi: 10.1093/bioinformatics/btq559 OpenUrl CrossRef PubMed Web of Science 32. ↵ Price AL , Weale ME , Patterson N , et al. ( 2008 ) Long-range LD can confound genome scans in admixed populations . Am J Hum Genet 83 ( 1 ): 132 – 135 ; author reply 135-139. doi: 10.1016/j.ajhg.2008.06.005 OpenUrl CrossRef PubMed Web of Science 33. ↵ Luo Y , Kanai M , Choi W , 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 ( 10 ): 1504 – 1516 . doi: 10.1038/s41588-021-00935-7 OpenUrl CrossRef PubMed 34. ↵ Robin X , Turck N , Hainard A , et al. ( 2011 ) pROC: an open-source package for R and S+ to analyze and compare ROC curves . BMC Bioinformatics 12 : 77 . doi: 10.1186/1471-2105-12-77 OpenUrl CrossRef PubMed 35. ↵ DeLong ER , DeLong DM , Clarke-Pearson DL ( 1988 ) Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach . Biometrics 44 ( 3 ): 837 – 845 OpenUrl CrossRef PubMed Web of Science 36. ↵ Sun X , Xu W ( 2014 ) Fast Implementation of DeLong’s Algorithm for Comparing the Areas Under Correlated Receiver Operating Characteristic Curves . IEEE Signal Process Lett 21 ( 11 ): 1389 – 1393 . doi: 10.1109/LSP.2014.2337313 OpenUrl CrossRef 37. ↵ Immel A , Key FM , Szolek A , et al. ( 2021 ) Analysis of Genomic DNA from Medieval Plague Victims Suggests Long-Term Effect of Yersinia pestis on Human Immunity Genes . Mol Biol Evol 38 ( 10 ): 4059 – 4076 . doi: 10.1093/molbev/msab147 OpenUrl CrossRef PubMed 38. ↵ Luckett AM , Weedon MN , Hawkes G , Leslie RD , Oram RA , Grant SFA ( 2023 ) Utility of genetic risk scores in type 1 diabetes . Diabetologia 66 ( 9 ): 1589 – 1600 . doi: 10.1007/s00125-023-05955-y OpenUrl CrossRef PubMed 39. ↵ Harrison JW , Tallapragada DSP , Baptist A , et al. ( 2020 ) Type 1 diabetes genetic risk score is discriminative of diabetes in non-Europeans: evidence from a study in India . Sci Rep 10 ( 1 ): 9450 . doi: 10.1038/s41598-020-65317-1 OpenUrl CrossRef PubMed 40. Qu H-Q , Qu J , Glessner J , et al. ( 2022 ) Improved genetic risk scoring algorithm for type 1 diabetes prediction . Pediatr Diabetes 23 ( 3 ): 320 – 323 . doi: 10.1111/pedi.13310 OpenUrl CrossRef PubMed 41. ↵ Deutsch AJ , Stalbow L , Majarian TD , et al. ( 2023 ) Polygenic Scores Help Reduce Racial Disparities in Predictive Accuracy of Automated Type 1 Diabetes Classification Algorithms . Diabetes Care 46 ( 4 ): 794 – 800 . doi: 10.2337/dc22-1833 OpenUrl CrossRef PubMed 42. ↵ Sankareswaran A , Kunte P , Fraser DP , et al. ( 2025 ) HLA haplotype diversity, islet autoantibody status and discriminative ability of type 1 diabetes genetic risk score in Indians . Diabet Med e70041 . doi: 10.1111/dme.70041 OpenUrl CrossRef 43. ↵ Luckett AM , Oram RA , Deutsch AJ , et al. ( 2025 ) Standardized Measurement of Type 1 Diabetes Polygenic Risk Across Multiancestry Population Cohorts . Diabetes Care 48 ( 6 ): e81 – e83 . doi: 10.2337/dc25-0142 OpenUrl CrossRef View the discussion thread. Back to top Previous Next Posted August 12, 2025. Download PDF Supplementary Material Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following HLA-focused type 1 diabetes genetic risk prediction in populations of diverse ancestry Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share HLA-focused type 1 diabetes genetic risk prediction in populations of diverse ancestry Dominika A. Michalek , Courtney Tern , Catherine C. Robertson , Wei-Min Chen , Suna Onengut-Gumuscu , Stephen S. Rich medRxiv 2025.08.07.25333167; doi: https://doi.org/10.1101/2025.08.07.25333167 Share This Article: Copy Citation Tools HLA-focused type 1 diabetes genetic risk prediction in populations of diverse ancestry Dominika A. Michalek , Courtney Tern , Catherine C. Robertson , Wei-Min Chen , Suna Onengut-Gumuscu , Stephen S. Rich medRxiv 2025.08.07.25333167; doi: https://doi.org/10.1101/2025.08.07.25333167 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Genetic and Genomic Medicine Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (300) Cardiovascular Medicine (4435) Dentistry and Oral Medicine (444) Dermatology (382) Emergency Medicine (608) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1509) Epidemiology (15228) Forensic Medicine (30) Gastroenterology (1124) Genetic and Genomic Medicine (6598) Geriatric Medicine (668) Health Economics (997) Health Informatics (4536) Health Policy (1368) Health Systems and Quality Improvement (1613) Hematology (540) HIV/AIDS (1264) Infectious Diseases (except HIV/AIDS) (15916) Intensive Care and Critical Care Medicine (1103) Medical Education (623) Medical Ethics (146) Nephrology (667) Neurology (6599) Nursing (346) Nutrition (998) Obstetrics and Gynecology (1144) Occupational and Environmental Health (957) Oncology (3332) Ophthalmology (974) Orthopedics (369) Otolaryngology (420) Pain Medicine (436) Palliative Medicine (130) Pathology (663) Pediatrics (1693) Pharmacology and Therapeutics (691) Primary Care Research (711) Psychiatry and Clinical Psychology (5447) Public and Global Health (9231) Radiology and Imaging (2198) Rehabilitation Medicine and Physical Therapy (1370) Respiratory Medicine (1196) Rheumatology (593) Sexual and Reproductive Health (712) Sports Medicine (530) Surgery (712) Toxicology (99) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a005a56afe540704',t:'MTc3OTU1NTkzMw=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();