GRAPHICAL ABSTRACT

Highlights

• This study explores sex differences in genetic risk for T2DM.

• Nine loci were identified with sex-dimorphic effects on the risk of T2DM.

• Sex-stratified and combined genetic scores were built and evaluated.

• Findings highlight the need for larger sex-specific studies to improve prediction.

INTRODUCTION

Diabetes is a growing global health challenge, with its prevalence expected to increase from 540 million individuals in 2021 to an estimated 783 million by 2045 [1]. Type 2 diabetes mellitus (T2DM), accounting for over 90% of these diabetes cases, is a multifactorial chronic condition influenced by both genetic and environmental factors and their interaction [2]. Genome-wide association studies (GWAS) have identified over 600 loci associated with T2DM risk [3]. Glycemic traits, such as fasting glucose (FG) and fasting insulin (FI), play a critical role in T2DM pathophysiology, and hundreds of variants associated with these traits have been identified in large-scale GWAS [4-8].

Polygenic risk scores (PRS), which aggregate the effects of multiple genetic variants, quantify an individual’s genetic susceptibility to diseases [9,10]. PRS can identify high-risk individuals who may not be classified accurately through clinical factors alone, aiding in early diagnosis and prevention efforts [11]. PRS can be constructed using methods such as P value thresholds followed by linkage disequilibrium (LD) pruning, which selects a subset of genetic variants from GWAS summary statistics [12], or shrinkage-based approaches that incorporate millions of variants while accounting for LD [13-15]. Recent efforts in PRS method development aim to improve prediction across diverse ancestries and integrate functional annotations to inform pathway-specific processes [16-19].

Most GWAS results used to construct PRS are from sex-combined analyses, assuming similar genetic effects between sexes. However, there is compelling evidence that T2DM genetic architecture may differ by sex. Women face greater relative risks of cardiovascular complications and mortality once diabetes develops [20,21]. Sex hormones and differences in fat distribution drive variation in glucose metabolism and insulin sensitivity [22,23]. In addition, recent studies have revealed sex-dimorphic effects in genetic associations for T2DM and glycemic traits. For example, T2DM risk signals near c-Maf inducing protein (CMIP), KLF transcription factor 14 (KLF14), and transcription factor 12 (TCF12) show larger effects in females, while the signal near ankyrin 1 (ANK1) displays a larger effect in males. FG signals at insulin receptor substrate 1 (IRS1) and zinc finger protein 12 (ZNF12) also showed sex-dimorphic effects [24,25].

Given these findings, we conducted a sex-specific T2DM GWAS meta-analysis using the latest genetic data from the UK Biobank (UKBB), Mass General Brigham Biobank (MGBB), and the Genetic Epidemiology Research on Adult Health and Aging (GERA) cohort. We also leveraged publicly available sex-specific GWAS summary statistics for FG and FI to derive PRS models and evaluated these models in 8,379 (1,303 cases and 7,076 controls) participants from the Framingham Heart Study (FHS). Our goal was to determine under what conditions sex-stratified PRS improve prediction accuracy over sex-combined PRS models for T2DM and glycemic traits and to assess the extent of any observed differences.

METHODS

Sex-specific and sex-combined GWAS

We utilized sex-specific and sex-combined GWAS summary statistics for T2DM, FG, and FI. Sex-specific T2DM GWAS were conducted across 41,444 cases (24,809 males and 16,635 females) and 354,539 controls (157,022 males and 197,517 females) using data from UKBB, MGBB, and GERA cohorts, all of predominantly European ancestry [26]. Genotype imputation was performed using the Trans-Omics for Precision Medicine (TOPMed) reference panel. Genome-wide genetic variants were tested for association using logistic regression models for binary T2DM outcome fitted using REGENIE [27], adjusting for age, body mass index (BMI), principal components (PCs) and imputation batch. Cohort-specific results were further meta-analyzed using METa AnaLysis Helper (METAL) [28].

For FG and FI, we obtained publicly available sex-specific GWAS results [24], including 67,506 males and 73,089 females for FG and 47,806 males and 50,404 females for FI across 73 studies. All participants were free of diabetes and of European ancestry. Each study performed sex-specific FG and FI GWAS under additive genetic models adjusting for age, study sites and PCs. Summary statistics were imputed for unmeasured variants using LD patterns from the 1000 Genomes reference panel via SSImp [29] and then meta-analyzed using Genome-Wide Association Meta-Analysis (GWAMA) [30]. We performed sex-combined GWAS by inverse variance weighting of sex-specific summary statistics using GWAMA.

We also conducted a two-step sex-heterogeneity test to identify variants with sex-dimorphic effects [30,31]. In the first step, we applied a 2-degree-of-freedom (df) test that jointly evaluates both the main genetic effect and the interaction with sex, using a genome-wide significance threshold (P<5×10^–8). In the second step, we tested for significant differences in effect sizes between sexes among the variants identified in step one. To correct for multiple testing, we used a Bonferroni-adjusted threshold based on the effective number of independent variants identified in step 1, estimated using the eigenvalue-based approach by Li and Ji [32].

PRS construction in the Framingham Heart Study

FHS participants were genotyped and imputed on the Michigan Imputation Server using Minimac3 with the Haplotype Reference Consortium (HRC, release 1.1, April 2016) reference panel [33]. Standard quality control was applied prior to imputation, including variant call rate ≥97%, Hardy–Weinberg equilibrium P≥1×10⁻⁶, <1,000 Mendelian errors, and minor allele frequency (MAF) ≥1%. Phasing was performed with SHAPEIT2 incorporating duoHMM to account for parental genotypes. Variants with imputation quality r²<0.3 were excluded.

Because FHS did not contribute to the GWAS for T2DM, FG, and FI, we developed PRS models using GWAS summary statistics and applied them to FHS participants. We used the PRS-continuous shrinkage (CS) method, which applies continuous shrinkage to estimate posterior effect sizes of genetic variants from GWAS summary statistics [14]. PRS-CS is a leading approach for constructing PRS in complex traits, as it leverages genome-wide LD structure, eliminates the need for arbitrary P value thresholds and LD pruning required by traditional approaches, and has demonstrated robust predictive performance across a variety of traits [34].

An external LD panel from the 1000 Genomes project was used for LD information in the PRS-CS [35], chosen to closely match the ancestry of the validation cohort. We calculated sex-stratified and sex-combined PRS, standardizing each to have a mean of zero and a standard deviation of one. Therefore, two sets of standardized PRS in FHS participants for each outcome: PRS[sex-stratified] and PRS[sex-combined].

PRS analysis in the Framingham Heart Study

Outcome definitions

We evaluated the associations of PRS with four outcomes: prevalent T2DM, incident T2DM, FG and FI in the FHS. T2DM status was determined based on physician diagnosis, self-reports, and diabetes medication use, and supplemented by glycemia measures (FG ≥7 mmol/L, glycosylated hemoglobin ≥6.5%, 2-hour oral glucose tolerance test ≥11.1 mmol/L, or random/non-FG ≥11.1 mmol/L). Individuals with pregnancy at diagnosis or age at diagnosis <25 years were excluded. Prevalent T2DM was defined using all available exams, and for incident T2DM analyses. Individuals with T2DM at baseline were excluded. Time-to-event was defined from the DNA draw date to the date of T2DM diagnosis or the end of followup, with a maximum follow-up of 26 years. For FG and FI analyses, we included only participants free of diabetes, using their baseline glycemic measures. FG was analyzed in mmol/L and FI was analyzed in pmol/L, with FI values log-transformed to reduce skewness.

PRS models

We compared PRS[sex-stratified] and PRS[sex-combined] models in females, males and all participants. For prevalent T2DM and the glycemic traits, we fit generalized mixed models (loglink for T2DM and linear for FG/FI) that incorporated the kinship matrix to account for familial relatedness. For incident T2DM, we used Cox proportional hazards model with robust sandwich estimators clustered by family ID to account for familial correlation.

All models were adjusted for age, sex, BMI, generation indicator, as FHS includes multiple generations [36,37], and PCs. The first 10 PCs were included in FG and FI analyses, and nominally significant PCs (P<0.05) were added in T2DM analyses. Odds ratios (ORs) were estimated for prevalent T2DM, hazard ratios (HRs) for incident T2DM, and standardized effect sizes for glycemic traits.

Comparison and evaluation

Model performances were evaluated in testing sets with 10-fold cross validation, where family members were kept intact to ensure familial independence across folds. We additionally constructed a PRS[mix] model defined as a weighted sum of PRS[sex-stratified] and PRS[sex-combined], with weights derived from a model including both sex-stratified and sex-combined PRS in the training set.

Evaluation metrics included the area under the receiver operating characteristic (AUC) for prevalent T2DM, Akaike information criterion for incident T2DM and R² increment (i.e., R² increase from null model with covariates only) for glycemic traits FG and FI. We further compared AUCs for prevalent T2DM models using Delong’s test, which evaluated whether AUC differences between models were statistically significant [38,39]. For incident T2DM, we performed partial likelihood to compare the performances of two PRS models.

Statistical analyses were performed using R version 4.2.1 (R Foundation for Statistical Computing, Vienna, Austria), and statistical significance was determined at a P value threshold of 0.05 unless otherwise noted. An overview of the study design and analytic workflow is provided in Fig. 1.

Ethics statement

This research was conducted using data and resources from the FHS of the National Heart Lung and Blood Institute of the National Institutes of Health and Boston University School of Medicine (Contract No. N01-HC-25195, HHSN268201500001I and 75N92019D00031). This study was approved by the Institutional Review Board (IRB) of the Boston University Medical Campus. All FHS participants provided written informed consent.

RESULTS

Sex-dimorphic effects at T2DM loci

We identified 12,651 variants that reached genome-wide significance (P<5×10⁻⁸) in the 2-df test which evaluates both main effects and their interaction with sex. Among these, nine T2DM-associated signals demonstrated significant sex-dimorphic effects, based on a Bonferroni correlation that accounted for the effective number of independent tests (0.05/739, heterogeneity P<6.8×10⁻⁵) with a MAF >0.0005 (Table 1, Supplementary Table 1, Supplementary Fig. 1). Several loci showed heterogeneity in association strengths or even opposite directions of effects between sexes, emphasizing the complexity of sex-dimorphic genetic regulation in T2DM.

The variant rs7524924 near lysophospholipase like 1 (LYPLAL1) showed a stronger protective effect in females than males (MAF=0.43, female OR=0.92, male OR=0.99, heterogeneity P=6.0×10⁻⁵). This variant is an adipose-specific expression quantitative trait locus (eQTL) for LYPLAL1-AS1 and has been associated with waist-hip ratio (WHR), a key obesity phenotype that influences T2DM risk in a sex-dimorphic manner [40].

Similarly, rs258494 near POC5 (MAF=0.44, female OR= 1.08, male OR=1.01, heterogeneity P=6.5×10^–5) and rs3923113 near growth factor receptor bound protein 14 (GRB14)/cordon-bleu WH2 repeat protein like 1 (COBLL1) (MAF=0.43, female OR=0.86, male OR=0.92, heterogeneity P=3.3×10^–5) also demonstrated stronger effects in females. As an adipose eQTL for COBLL1, rs3923113 has been associated with insulin resistance and FI levels, and showed sex-dimorphic effects on WHR trait [41].

We also identified rare variants with significant sex-dimorphic effects. Allele A of rs141203601, located upstream of NK6 homeobox 1 (NKX6-1), showed a highly protective effect in females but was associated with increased T2DM risk in males (MAF=0.0010, female OR=0.20, male OR=6.92, heterogeneity P=1.1×10^–8). Another rare variant, rs146239492 near kinesin family member 2B (KIF2B), also showed significant sex-dimorphic effects (MAF=0.0031, female OR=0.43, male OR=5.20, heterogeneity P=3.8×10^–8).

The variant rs73795193 near fibroblast growth factor 1 (FGF1) showed an increased risk in females but not in males (MAF=0.04, female OR=1.26, male OR=0.96, heterogeneity P=5.2×10^–7). Though this variant has limited evidence of association with T2DM, FGF1 plays a key role in glucose homeostasis and adipogenesis, which may explain its involvement in sex-specific metabolic regulation [42].

Other loci with significant sex-heterogeneous effects included KLF14 (rs59326756, MAF=0.29, female OR=1.15, male OR=1.05, heterogeneity P=4.1×10^–7), phosphodiesterase 3A (PDE3A) (rs7134375, MAF=0.39, female OR=0.93, male OR=1.00, heterogeneity P=2.7×10^–5), and CMIP (rs12443634, MAF=0.27, female OR=0.87, male OR=0.96, heterogeneity P=2.6×10^–7).

We did not investigate sex-heterogenous variants for FG and FI, as these variants were previously reported [24].

Participant characteristics in FHS

This study includes 8,379 FHS participants across three generational cohorts. All participants are of European ancestry. The characteristics of these participants are summarized in Table 2.

PRS of T2DM, FG, and FI

The correlation between PRS[sex-stratified] and PRS[sex-combined]) was higher for T2DM than for FG and FI (Pearson correlation: ρ_T2D=0.82, ρ_FG=0.77 and ρ_FI=0.66) (Supplementary Fig. 2). Associations between PRS and T2DM, FG, and FI were assessed separately for three groups: (1) all participants; (2) males; and (3) females. For prevalent T2DM analysis in all participants, OR was 1.67 per each standard deviation for PRS [sex-stratified] (95% CI for OR, 1.56 to 1.79) compared to 1.87 for PRS[sex-combined] (95% CI for OR, 1.74 to 2.01). For incident T2DM in all participants, HR was 1.45 per each standard deviation for PRS[sex-stratified] (95% CI for HR, 1.31 to 1.60) and 1.58 for PRS[sex-combined] (95% CI for HR, 1.42 to 1.76). Although these effect sizes were numerically different, overlapping confidence intervals for ORs and HRs indicated no statistically significant differences between the models. We observed similar overlapping patterns between PRS[sex-stratified] and PRS[sex-combined] for FG and log-transformed FI (Fig. 2).

Combined PRS prediction on T2DM, FG, and FI

We then developed a model, called PRS[mix], that combines both sex-stratified and sex-combined PRS using a training dataset from FHS and tests it in an independent FHS subset of participants. For the combined model PRS[mix], it consistently assigned less than 25% weight to PRS[sex-stratified] across all outcomes. The contribution of PRS[sex-stratified] was relatively higher for glycemic traits (FG and FI) than for T2DM outcomes (Supplementary Fig. 3).

Across all four outcomes, PRS[sex-combined] constantly showed the best predictive accuracy. This may suggest that sex-combined PRS model captures enough variability without sex-stratification (Table 3). For prevalent T2DM, Delong’s test showed significant differences between PRS[sex-stratified] and both PRS[sex-combined] and PRS[mix] models (P=9.6×10⁻⁷ and P=2.1×10⁻⁷). In incident T2DM analysis, partial likelihood ratio tests indicated that both PRS[sex-combined] and PRS[mix] models provided significantly different fits compared to the PRS[sex-stratified] model (P=0.017 and P=0.013, respectively) (Table 4).

DISCUSSION

This study investigates the role of sex-dimorphic genetic effects in the identification of new variants associated with T2DM, and the development of PRS models for prediction of T2DM and glycemic traits (FG and FI). Among the identified T2DM loci, variant rs12443634 at CMIP locus presented a greater risk effect in females than in males. We also observed a similar pattern for rs2925979, a variant in high LD (R²>0.8) with variant rs12443634, which showed a modestly increased risk in females but not in males, supporting a stronger female-specific effect at this locus (Supplementary Table 1). Notably, rs2925979 was identified with significant sex-dimorphic effects in a previous large-scale T2DM GWAS meta-analysis [25]. In a Chinese family-based study, rs2925979 T allele was associated with 29% higher odds of T2DM in females but did not show a significant association in males [43]. Interestingly, their study found this allele was also associated with lower BMI and body fat percentage in non-diabetic females, suggesting pleiotropic effects of CMIP on both T2DM and obesity-related phenotypes. These findings underscore the importance of further investigation of the independent mechanisms underlying CMIP’s sex-dimorphic effects. In addition to CMIP region, we confirmed sex-dimorphic effects near GRB14/COBLL1 and KLF14, which only showed nominally differences in risk effects in a previous study [25].

Novel loci with sex-dimorphic effects were also identified, including LYPLAL1, NKX6-1, POC5, FGF1, PDE3A, and KIF2B. These findings contribute to the growing evidence of sex-specific genetic architecture in T2DM. LYPLAL1 has been implicated in lipid metabolism and fat distribution, and previous studies supported the sex-specific mechanism at this locus [44,45].

NKX6-1 is a transcription factor essential for pancreatic β-cell function, and its dysregulation may contribute to divergent T2DM risk between sexes [46]. FGF1 plays a role in adipogenesis and glucose homeostasis, with evidence suggesting differential gene expression by sex [47]. Understanding the biological relevance of these findings could provide insight into how sex-specific regulation of gene expression contributes to metabolic differences between males and females.

Our findings reveal that PRS derived from sex-combined GWAS better predict outcomes than those derived from sex-stratified GWAS for the outcomes investigated. These results suggest that current sample sizes in sex-stratified GWAS of T2DM, FG, and FI may be insufficient to detect subtle genetic contributions and improve predictive accuracy relative to sex-combined models.

Reduced statistical power is a key limitation of sex-stratified models. Despite identifying nine loci with significant sex-dimorphic effects for T2DM, the sample size in each sex-specific GWAS may have been too small to capture additional associations. Although the T2DM GWAS used to derive PRS were conducted in very large cohorts (UKBB, MGBB, and GERA), stratifying by sex effectively reduces the sample size within each analysis. This reduced power likely explains the weaker performance of sex-stratified PRS compared to sex-combined PRS. As larger sex-specific GWAS become available in the future, sex-stratified PRS may capture additional sex-dimorphic effects, and their predictive performance could improve. Consequently, sex-stratified PRS may not fully capture genetic risk variation, while sex-combined PRS can leverage shared genetic effects across sexes for greater detection power. Moreover, a relatively small number of variants with significant sex-dimorphic effects for the outcomes investigated limits the ability of sex-stratified PRS to demonstrate a performance advantage. Therefore, using sex-combined PRS with larger sample sizes remains advantageous in current research settings for T2DM, FG, and FI. Future more powerful sex-specific GWAS should identify additional sex-specific effects and may improve the performance of sex-dimorphic effects.

Another interesting observation was that PRS[sex-stratified] contributed more to PRS[mix] model for glycemic traits than for T2DM outcomes. This may reflect stronger sex-specific genetic influences on intermediate glycemic traits, which allow sex-stratified PRS to capture additional variance that contributes to prediction accuracy. Future studies could investigate how to best optimize the integration of sex-specific and sex-combined PRS across different traits and outcomes.

From a clinical perspective, PRS are valuable tools for identifying individuals at high genetic risk for diseases like T2DM, particularly among those perceived to have low risk based on classic clinical risk factors [11]. However, our results suggest that current sex-specific PRS models offer limited additional predictive value over sex-combined models for T2DM and glycemic traits. This finding highlights the limitation of current sex-stratified GWAS sample sizes, underscores the need for enlarging such datasets to enhance precision and cautions against premature clinical application until then. Future research should aim to increase the sample sizes of sex-stratified GWAS, particularly in non-European populations. This would improve the power to detect sex-dimorphic genetic effects and refine PRS models. Additionally, integrating multi-omics data and functional annotations may provide further insights into sex-specific mechanisms underlying T2DM and related traits. Encouragingly, studies of other complex traits, such as schizophrenia, coronary artery disease, and basal cell carcinoma, have demonstrated the clinical utility of sex-stratified PRS for risk stratification, supporting the need for continued research in this area [48-50].

In conclusion, while sex-combined PRS currently offer better predictive performance for T2DM and glycemic traits, ongoing research into sex-dimorphic effects remains crucial for advancing personalized medicine and improving disease risk prediction. Expanding sample sizes and incorporating functional insights will be essential for optimizing the integration of sex-stratified models into clinical practice.

SUPPLEMENTARY MATERIALS

NOTES

CONFLICTS OF INTEREST

No potential conflict of interest relevant to this article was reported.

AUTHOR CONTRIBUTIONS

Conception or design: N.W., Y.Z., J.B.M., A.K.M., C.T.L., J.D., J.M.M.

Acquisition, analysis, or interpretation of data: all authors.

Drafting the work or revising: N.W., Y.Z., C.T.L., J.D., J.M.M.

Final approval of the manuscript: all authors.

FUNDING

Ningyuan Wang, Yixin Zhang, James B. Meigs, Alisa K. Manning, Ching-Ti Liu, Josée Dupuis are supported by NIDDK UM1DK078616. Josep M. Mercader is supported by American Diabetes Association grant #11-22-ICTSPM-16 and by NHGRI U01HG011723, by the National Institute Of Diabetes And Digestive And Kidney Diseases of the National Institutes of Health under Award Number R01DK137993 and U01 DK140757, AMP CMD award from RFP 6 from the Foundation for the National Institutes of Health, and a Medical University of Bialystok (MUB) grant from the Ministry of Science and Higher Education (Poland). This work is supported by the Novo Nordisk Foundation (NNF21SA0072102).

ACKNOWLEDGMENTS

The authors thank the participants of the Framingham Heart Study, as well as the study team for their contributions.

Fig. 1.

Study design and analytic workflow. Sex-specific and sex-combined type 2 diabetes mellitus (T2DM) genome-wide association studies (GWAS) were conducted in the UK Biobank (UKBB), Mass General Brigham Biobank (MGBB), and Genetic Epidemiology Research on Adult Health and Aging (GERA) cohorts. Fasting glucose (FG) and fasting insulin (FI) sex-specific and sex-combined GWAS results were obtained from The Malignant Germ Cell International Consortium (MaGIC) [24]. A two-step heterogeneity test was applied to identify variants with sex-dimorphic effects. Polygenic risk score (PRS) were constructed in the Framingham Heart Study (FHS) using PRS-continuous shrinkage (CS) with both sex-specific and sex-combined GWAS as input, along with the 1000 Genomes European linkage disequilibrium (LD) panel. In the FHS validation cohort, PRS were evaluated for four outcomes: prevalent T2DM, incident T2DM, FG, and FI. Mixed effects models and Cox proportional hazards model accounting for familial relatedness were used, adjusting for age, body mass index (BMI), generation, and principal components (PCs). Model performances were assessed with 10-fold cross validation using area under the receiver operating characteristic (AUC), Akaike information criterion (AIC), and R² increment. GWAMA, Genome-Wide Association Meta-Analysis; OR, odds ratio; HR, hazard ratio. ^aSex-specific GWAS summary statistics are publicly available from Lagou et al. [24].

Fig. 2.

Effects of two polygenic risk score (PRS) on prevalent type 2 diabetes mellitus (T2DM), incident T2DM, fasting glucose (FG), and fasting insulin (FI). Forest plots show the effect sizes of PRS[sex-stratified] and PRS[sex-combined] models for all participants, females and males across four outcomes. Odds ratios (ORs) are presented for prevalent T2DM, hazard ratios (HRs) for incident T2DM, and regression coefficient (beta) for FG and log-transformed FI. Effect sizes represent changes per 1 standard deviation increase in the PRS. Horizontal lines indicate 95% confidence interval (CI) for each effect size.

REFERENCES

• 1. Magliano DJ, Boyko EJ. IDF Diabetes Atlas. 10th ed. Brussels: International Diabetes Federation; 2021.
• 2. Murea M, Ma L, Freedman BI. Genetic and environmental factors associated with type 2 diabetes and diabetic vascular complications. Rev Diabet Stud 2012;9:6-22.ArticlePubMedPMC
• 3. Suzuki K, Hatzikotoulas K, Southam L, Taylor HJ, Yin X, Lorenz KM, et al. Genetic drivers of heterogeneity in type 2 diabetes pathophysiology. Nature 2024;627:347-57.PubMedPMC
• 4. Dupuis J, Langenberg C, Prokopenko I, Saxena R, Soranzo N, Jackson AU, et al. New genetic loci implicated in fasting glucose homeostasis and their impact on type 2 diabetes risk. Nat Genet 2010;42:105-16.PubMedPMC
• 5. Lagou V, Jiang L, Ulrich A, Zudina L, Gonzalez KS, Balkhiyarova Z, et al. GWAS of random glucose in 476,326 individuals provide insights into diabetes pathophysiology, complications and treatment stratification. Nat Genet 2023;55:1448-61.PubMedPMC
• 6. Manning AK, Hivert MF, Scott RA, Grimsby JL, Bouatia-Naji N, Chen H, et al. A genome-wide approach accounting for body mass index identifies genetic variants influencing fasting glycemic traits and insulin resistance. Nat Genet 2012;44:659-69.PubMedPMC
• 7. Scott RA, Lagou V, Welch RP, Wheeler E, Montasser ME, Luan J, et al. Large-scale association analyses identify new loci influencing glycemic traits and provide insight into the underlying biological pathways. Nat Genet 2012;44:991-1005.PubMedPMC
• 8. Chen J, Spracklen CN, Marenne G, Varshney A, Corbin LJ, Luan J, et al. The trans-ancestral genomic architecture of glycemic traits. Nat Genet 2021;53:840-60.PubMedPMC
• 9. Ge T, Irvin MR, Patki A, Srinivasasainagendra V, Lin YF, Tiwari HK, et al. Development and validation of a trans-ancestry polygenic risk score for type 2 diabetes in diverse populations. Genome Med 2022;14:70.PubMedPMC
• 10. Mercader JM, Ng MC, Manning AK, Rich SS. Predicting diabetes risk in diverse populations: what next? Lancet Diabetes Endocrinol 2021;9:808-10.ArticlePubMedPMC
• 11. Mandla R, Schroeder P, Porneala B, Florez JC, Meigs JB, Mercader JM, et al. Polygenic scores for longitudinal prediction of incident type 2 diabetes in an ancestrally and medically diverse primary care physician network: a patient cohort study. Genome Med 2024;16:63.ArticlePubMedPMCPDF
• 12. Purcell SM, Wray NR, Stone JL, Visscher PM, O’Donovan MC, Sullivan PF, et al. Common polygenic variation contributes to risk of schizophrenia and bipolar disorder. Nature 2009;460:748-52.ArticlePubMedPMCPDF
• 13. Chun S, Imakaev M, Hui D, Patsopoulos NA, Neale BM, Kathiresan S, et al. Non-parametric polygenic risk prediction via partitioned GWAS summary statistics. Am J Hum Genet 2020;107:46-59.ArticlePubMedPMC
• 14. Ge T, Chen CY, Ni Y, Feng YA, Smoller JW. Polygenic prediction via Bayesian regression and continuous shrinkage priors. Nat Commun 2019;10:1776.ArticlePubMedPMCPDF
• 15. Vilhjalmsson BJ, Yang J, Finucane HK, Gusev A, Lindstrom S, Ripke S, et al. Modeling linkage disequilibrium increases accuracy of polygenic risk scores. Am J Hum Genet 2015;97:576-92.ArticlePubMedPMC
• 16. Kachuri L, Chatterjee N, Hirbo J, Schaid DJ, Martin I, Kullo IJ, et al. Principles and methods for transferring polygenic risk scores across global populations. Nat Rev Genet 2024;25:8-25.ArticlePubMedPDF
• 17. Ruan Y, Lin YF, Feng YA, Chen CY, Lam M, Guo Z, et al. Improving polygenic prediction in ancestrally diverse populations. Nat Genet 2022;54:573-80.PubMedPMC
• 18. Smith K, Deutsch AJ, McGrail C, Kim H, Hsu S, Huerta-Chagoya A, et al. Multi-ancestry polygenic mechanisms of type 2 diabetes. Nat Med 2024;30:1065-74.ArticlePubMedPMCPDF
• 19. Udler MS, Kim J, von Grotthuss M, Bonas-Guarch S, Cole JB, Chiou J, et al. Type 2 diabetes genetic loci informed by multitrait associations point to disease mechanisms and subtypes: a soft clustering analysis. PLoS Med 2018;15:e1002654.ArticlePubMedPMC
• 20. Kautzky-Willer A, Leutner M, Harreiter J. Sex differences in type 2 diabetes. Diabetologia 2023;66:986-1002.ArticlePubMedPMCPDF
• 21. Wang Y, O'Neil A, Jiao Y, Wang L, Huang J, Lan Y, et al. Sex differences in the association between diabetes and risk of cardiovascular disease, cancer, and all-cause and cause-specific mortality: a systematic review and meta-analysis of 5,162,654 participants. BMC Med 2019;17:136.ArticlePubMedPMCPDF
• 22. Mauvais-Jarvis F. Gender differences in glucose homeostasis and diabetes. Physiol Behav 2018;187:20-3.ArticlePubMed
• 23. Kautzky-Willer A, Harreiter J, Pacini G. Sex and gender differences in risk, pathophysiology and complications of type 2 diabetes mellitus. Endocr Rev 2016;37:278-316.ArticlePubMedPMC
• 24. Lagou V, Magi R, Hottenga JJ, Grallert H, Perry JRB, Bouatia-Naji N, et al. Sex-dimorphic genetic effects and novel loci for fasting glucose and insulin variability. Nat Commun 2021;12:24.PubMedPMC
• 25. Mahajan A, Taliun D, Thurner M, Robertson NR, Torres JM, Rayner NW, et al. Fine-mapping type 2 diabetes loci to single-variant resolution using high-density imputation and islet-specific epigenome maps. Nat Genet 2018;50:1505-13.PubMedPMC
• 26. Huerta-Chagoya A, Schroeder P, Mandla R, Li J, Morris L, Vora M, et al. Rare variant analyses in 51,256 type 2 diabetes cases and 370,487 controls reveal the pathogenicity spectrum of monogenic diabetes genes. Nat Genet 2024;56:2370-9.ArticlePubMedPMCPDF
• 27. Mbatchou J, Barnard L, Backman J, Marcketta A, Kosmicki JA, Ziyatdinov A, et al. Computationally efficient whole-genome regression for quantitative and binary traits. Nat Genet 2021;53:1097-103.ArticlePubMedPDF
• 28. Willer CJ, Li Y, Abecasis GR. METAL: fast and efficient meta-analysis of genomewide association scans. Bioinformatics 2010;26:2190-1.ArticlePubMedPMCPDF
• 29. Rueger S, McDaid A, Kutalik Z. Evaluation and application of summary statistic imputation to discover new height-associated loci. PLoS Genet 2018;14:e1007371.ArticlePubMedPMC
• 30. Magi R, Morris AP. GWAMA: software for genome-wide association meta-analysis. BMC Bioinformatics 2010;11:288.PubMedPMC
• 31. Magi R, Lindgren CM, Morris AP. Meta-analysis of sex-specific genome-wide association studies. Genet Epidemiol 2010;34:846-53.ArticlePubMedPMC
• 32. Li J, Ji L. Adjusting multiple testing in multilocus analyses using the eigenvalues of a correlation matrix. Heredity (Edinb) 2005;95:221-7.ArticlePubMedPDF
• 33. McCarthy S, Das S, Kretzschmar W, Delaneau O, Wood AR, Teumer A, et al. A reference panel of 64,976 haplotypes for genotype imputation. Nat Genet 2016;48:1279-83.ArticlePubMedPMCPDF
• 34. Pain O, Glanville KP, Hagenaars SP, Selzam S, Furtjes AE, Gaspar HA, et al. Evaluation of polygenic prediction methodology within a reference-standardized framework. PLoS Genet 2021;17:e1009021.ArticlePubMedPMC
• 35. Auton A, Brooks LD, Durbin RM, Garrison EP, Kang HM, Korbel JO, et al. A global reference for human genetic variation. Nature 2015;526:68-74.PubMedPMC
• 36. Breslow NE, Clayton DG. Approximate inference in generalized linear mixed models. J Am Stat Assoc 1993;88:9-25.Article
• 37. Chen H, Wang C, Conomos MP, Stilp AM, Li Z, Sofer T, et al. Control for population structure and relatedness for binary traits in genetic association studies via logistic mixed models. Am J Hum Genet 2016;98:653-66.ArticlePubMedPMC
• 38. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988;44:837-45.ArticlePubMed
• 39. Sun X, Xu W. Fast implementation of DeLong’s algorithm for comparing the areas under correlated receiver operating characteristic curves. IEEE Signal Process Lett 2014;21:1389-93.Article
• 40. Heid IM, Jackson AU, Randall JC, Winkler TW, Qi L, Steinthorsdottir V, et al. Meta-analysis identifies 13 new loci associated with waist-hip ratio and reveals sexual dimorphism in the genetic basis of fat distribution. Nat Genet 2010;42:949-60.PubMedPMC
• 41. Glunk V, Laber S, Sinnott-Armstrong N, Sobreira DR, Strobel SM, Batista TM, et al. A non-coding variant linked to metabolic obesity with normal weight affects actin remodelling in subcutaneous adipocytes. Nat Metab 2023;5:861-79.ArticlePubMedPMCPDF
• 42. Nies VJ, Sancar G, Liu W, van Zutphen T, Struik D, Yu RT, et al. Fibroblast growth factor signaling in metabolic regulation. Front Endocrinol (Lausanne) 2015;6:193.ArticlePubMed
• 43. Cao Y, Wang T, Wu Y, Juan J, Qin X, Tang X, et al. Opposite genetic effects of CMIP polymorphisms on the risk of type 2 diabetes and obesity: a family-based study in China. Int J Mol Sci 2018;19:1011.ArticlePubMedPMC
• 44. Vohnoutka RB, Kuppa A, Hegde Y, Chen Y, Pant A, Tohme ME, et al. Knockout of murine Lyplal1 confers sex-specific protection against diet-induced obesity. J Mol Endocrinol 2023;70:e220131.ArticlePubMedPMC
• 45. Lei X, Callaway M, Zhou H, Yang Y, Chen W. Obesity associated Lyplal1 gene is regulated in diet induced obesity but not required for adipocyte differentiation. Mol Cell Endocrinol 2015;411:207-13.ArticlePubMed
• 46. Aigha II, Abdelalim EM. NKX6.1 transcription factor: a crucial regulator of pancreatic β cell development, identity, and proliferation. Stem Cell Res Ther 2020;11:459.ArticlePubMedPMCPDF
• 47. Schneider JS, Anderson DW, Sonnenahalli H, Vadigepalli R. Sex-based differences in gene expression in hippocampus following postnatal lead exposure. Toxicol Appl Pharmacol 2011;256:179-90.ArticlePubMedPMC
• 48. Mas-Bermejo P, Papiol S, Torrecilla P, Lavin V, Kwapil TR, Barrantes-Vidal N, et al. Sex-specific association between schizophrenia polygenic risk and subclinical schizophrenia-related traits. Prog Neuropsychopharmacol Biol Psychiatry 2025;136:111161.ArticlePubMed
• 49. Paruchuri K, Bhukar R, Urbut S, Ruan Y, Ganesh S, Postupaka D, et al. Using sex-specific polygenic risk to prognosticate coronary artery disease in women. J Am Heart Assoc 2024;13:e034946.ArticlePubMedPMC
• 50. Roberts MR, Sordillo JE, Kraft P, Asgari MM. Sex-stratified polygenic risk score identifies individuals at increased risk of basal cell carcinoma. J Invest Dermatol 2020;140:971-5.ArticlePubMed

Figure & Data

References

Citations

Citations to this article as recorded by  

About this article

Wang N, Zhang Y, Schroeder P, Huerta-Chagoya A, Mandla R, Meigs JB, Manning AK, Liu CT, Dupuis J, Mercader JM. Evaluation of Sex-Stratified Polygenic Risk Scores for Type 2 Diabetes Mellitus and Glycemic Traits in the Framingham Heart Study. Diabetes Metab J. 2025 Dec 9. doi: 10.4093/dmj.2025.0557. Epub ahead of print.

Received: Jun 25, 2025; Accepted: Oct 14, 2025

DOI: https://doi.org/10.4093/dmj.2025.0557.

Download citation ↓