Toward a fine-scale population health monitoring system.

Understanding population health disparities is an essential component of equitable precision health efforts. Epidemiology research often relies on definitions of race and ethnicity, but these population labels may not adequately capture disease burdens and environmental factors impacting specific sub-populations. Here, we propose a framework for repurposing data from electronic health records (EHRs) in concert with genomic data to explore the demographic ties that can impact disease burdens. Using data from a diverse biobank in New York City, we identified 17 communities sharing recent genetic ancestry. We observed 1,177 health outcomes that were statistically associated with a specific group and demonstrated significant differences in the segregation of genetic variants contributing to Mendelian diseases. We also demonstrated that fine-scale population structure can impact the prediction of complex disease risk within groups. This work reinforces the utility of linking genomic data to EHRs and provides a framework toward fine-scale monitoring of population health.

Predicted gene expression in ancestrally diverse populations leads to discovery of susceptibility loci for lifestyle and cardiometabolic traits.

One mechanism by which genetic factors influence complex traits and diseases is altering gene expression. Direct measurement of gene expression in relevant tissues is rarely tenable; however, genetically regulated gene expression (GReX) can be estimated using prediction models derived from large multi-omic datasets. These approaches have led to the discovery of many gene-trait associations, but whether models derived from predominantly European ancestry (EA) reference panels can map novel associations in ancestrally diverse populations remains unclear. We applied PrediXcan to impute GReX in 51,520 ancestrally diverse Population Architecture using Genomics and Epidemiology (PAGE) participants (35% African American, 45% Hispanic/Latino, 10% Asian, and 7% Hawaiian) across 25 key cardiometabolic traits and relevant tissues to identify 102 novel associations. We then compared associations in PAGE to those in a random subset of 50,000 White British participants from UK Biobank (UKBB50k) for height and body mass index (BMI). We identified 517 associations across 47 tissues in PAGE but not UKBB50k, demonstrating the importance of diverse samples in identifying trait-associated GReX. We observed that variants used in PrediXcan models were either more or less differentiated across continental-level populations than matched-control variants depending on the specific population reflecting sampling bias. Additionally, variants from identified genes specific to either PAGE or UKBB50k analyses were more ancestrally differentiated than those in genes detected in both analyses, underlining the value of population-specific discoveries. This suggests that while EA-derived transcriptome imputation models can identify new associations in non-EA populations, models derived from closely matched reference panels may yield further insights. Our findings call for more diversity in reference datasets of tissue-specific gene expression.

Genetic analyses of diverse populations improves discovery for complex traits.

Genome-wide association studies (GWAS) have laid the foundation for investigations into the biology of complex traits, drug development and clinical guidelines. However, the majority of discovery efforts are based on data from populations of European ancestry1-3. In light of the differential genetic architecture that is known to exist between populations, bias in representation can exacerbate existing disease and healthcare disparities. Critical variants may be missed if they have a low frequency or are completely absent in European populations, especially as the field shifts its attention towards rare variants, which are more likely to be population-specific4-10. Additionally, effect sizes and their derived risk prediction scores derived in one population may not accurately extrapolate to other populations11,12. Here we demonstrate the value of diverse, multi-ethnic participants in large-scale genomic studies. The Population Architecture using Genomics and Epidemiology (PAGE) study conducted a GWAS of 26 clinical and behavioural phenotypes in 49,839 non-European individuals. Using strategies tailored for analysis of multi-ethnic and admixed populations, we describe a framework for analysing diverse populations, identify 27 novel loci and 38 secondary signals at known loci, as well as replicate 1,444 GWAS catalogue associations across these traits. Our data show evidence of effect-size heterogeneity across ancestries for published GWAS associations, substantial benefits for fine-mapping using diverse cohorts and insights into clinical implications. In the United States-where minority populations have a disproportionately higher burden of chronic conditions13-the lack of representation of diverse populations in genetic research will result in inequitable access to precision medicine for those with the highest burden of disease. We strongly advocate for continued, large genome-wide efforts in diverse populations to maximize genetic discovery and reduce health disparities.

Minority-centric meta-analyses of blood lipid levels identify novel loci in the Population Architecture using Genomics and Epidemiology (PAGE) study.

Lipid levels are important markers for the development of cardio-metabolic diseases. Although hundreds of associated loci have been identified through genetic association studies, the contribution of genetic factors to variation in lipids is not fully understood, particularly in U.S. minority groups. We performed genome-wide association analyses for four lipid traits in over 45,000 ancestrally diverse participants from the Population Architecture using Genomics and Epidemiology (PAGE) Study, followed by a meta-analysis with several European ancestry studies. We identified nine novel lipid loci, five of which showed evidence of replication in independent studies. Furthermore, we discovered one novel gene in a PrediXcan analysis, minority-specific independent signals at eight previously reported loci, and potential functional variants at two known loci through fine-mapping. Systematic examination of known lipid loci revealed smaller effect estimates in African American and Hispanic ancestry populations than those in Europeans, and better performance of polygenic risk scores based on minority-specific effect estimates. Our findings provide new insight into the genetic architecture of lipid traits and highlight the importance of conducting genetic studies in diverse populations in the era of precision medicine.

EPS: automated feature selection in case-control studies using extreme pseudo-sampling.

SUMMARY: Finding informative predictive features in high-dimensional biological case-control datasets is challenging. The Extreme Pseudo-Sampling (EPS) algorithm offers a solution to the challenge of feature selection via a combination of deep learning and linear regression models. First, using a variational autoencoder, it generates complex latent representations for the samples. Second, it classifies the latent representations of cases and controls via logistic regression. Third, it generates new samples (pseudo-samples) around the extreme cases and controls in the regression model. Finally, it trains a new regression model over the upsampled space. The most significant variables in this regression are selected. We present an open-source implementation of the algorithm that is easy to set up, use and customize. Our package enhances the original algorithm by providing new features and customizability for data preparation, model training and classification functionalities. We believe the new features will enable the adoption of the algorithm for a diverse range of datasets. AVAILABILITY AND IMPLEMENTATION: The software package for Python is available online at https://github.com/roohy/eps. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

W-TSS: A Wavelet-Based Algorithm for Discovering Time Series Shapelets.

Many approaches to time series classification rely on machine learning methods. However, there is growing interest in going beyond black box prediction models to understand discriminatory features of the time series and their associations with outcomes. One promising method is time-series shapelets (TSS), which identifies maximally discriminative subsequences of time series. For example, in environmental health applications TSS could be used to identify short-term patterns in exposure time series (shapelets) associated with adverse health outcomes. Identification of candidate shapelets in TSS is computationally intensive. The original TSS algorithm used exhaustive search. Subsequent algorithms introduced efficiencies by trimming/aggregating the set of candidates or training candidates from initialized values, but these approaches have limitations. In this paper, we introduce Wavelet-TSS (W-TSS) a novel intelligent method for identifying candidate shapelets in TSS using wavelet transformation discovery. We tested W-TSS on two datasets: (1) a synthetic example used in previous TSS studies and (2) a panel study relating exposures from residential air pollution sensors to symptoms in participants with asthma. Compared to previous TSS algorithms, W-TSS was more computationally efficient, more accurate, and was able to discover more discriminative shapelets. W-TSS does not require pre-specification of shapelet length.

Discovery, fine-mapping, and conditional analyses of genetic variants associated with C-reactive protein in multiethnic populations using the Metabochip in the Population Architecture using Genomics and Epidemiology (PAGE) study.

C-reactive protein (CRP) is a circulating biomarker indicative of systemic inflammation. We aimed to evaluate genetic associations with CRP levels among non-European-ancestry populations through discovery, fine-mapping and conditional analyses. A total of 30 503 non-European-ancestry participants from 6 studies participating in the Population Architecture using Genomics and Epidemiology study had serum high-sensitivity CRP measurements and ∼200 000 single nucleotide polymorphisms (SNPs) genotyped on the Metabochip. We evaluated the association between each SNP and log-transformed CRP levels using multivariate linear regression, with additive genetic models adjusted for age, sex, the first four principal components of genetic ancestry, and study-specific factors. Differential linkage disequilibrium patterns between race/ethnicity groups were used to fine-map regions associated with CRP levels. Conditional analyses evaluated for multiple independent signals within genetic regions. One hundred and sixty-three unique variants in 12 loci in overall or race/ethnicity-stratified Metabochip-wide scans reached a Bonferroni-corrected P-value <2.5E-7. Three loci have no (HACL1, OLFML2B) or only limited (PLA2G6) previous associations with CRP levels. Six loci had different top hits in race/ethnicity-specific versus overall analyses. Fine-mapping refined the signal in six loci, particularly in HNF1A. Conditional analyses provided evidence for secondary signals in LEPR, IL1RN and HNF1A, and for multiple independent signals in CRP and APOE. We identified novel variants and loci associated with CRP levels, generalized known CRP associations to a multiethnic study population, refined association signals at several loci and found evidence for multiple independent signals at several well-known loci. This study demonstrates the benefit of conducting inclusive genetic association studies in large multiethnic populations.

Northwestern University schizophrenia data sharing for SchizConnect: A longitudinal dataset for large-scale integration.

In this paper, we describe an instance of the Northwestern University Schizophrenia Data and Software Tool (NUSDAST), a schizophrenia-related dataset hosted at XNAT Central, and the SchizConnect data portal used for accessing and sharing the dataset. NUSDAST was built and extended upon existing, standard schemas available for data sharing on XNAT Central (http://central.xnat.org/). With the creation of SchizConnect, we were able to link NUSDAST to other neuroimaging data sources and create a powerful, federated neuroimaging resource.

SchizConnect: Mediating neuroimaging databases on schizophrenia and related disorders for large-scale integration.

SchizConnect (www.schizconnect.org) is built to address the issues of multiple data repositories in schizophrenia neuroimaging studies. It includes a level of mediation--translating across data sources--so that the user can place one query, e.g. for diffusion images from male individuals with schizophrenia, and find out from across participating data sources how many datasets there are, as well as downloading the imaging and related data. The current version handles the Data Usage Agreements across different studies, as well as interpreting database-specific terminologies into a common framework. New data repositories can also be mediated to bring immediate access to existing datasets. Compared with centralized, upload data sharing models, SchizConnect is a unique, virtual database with a focus on schizophrenia and related disorders that can mediate live data as information is being updated at each data source. It is our hope that SchizConnect can facilitate testing new hypotheses through aggregated datasets, promoting discovery related to the mechanisms underlying schizophrenic dysfunction.

Phenome-wide association study (PheWAS) for detection of pleiotropy within the Population Architecture using Genomics and Epidemiology (PAGE) Network.

Using a phenome-wide association study (PheWAS) approach, we comprehensively tested genetic variants for association with phenotypes available for 70,061 study participants in the Population Architecture using Genomics and Epidemiology (PAGE) network. Our aim was to better characterize the genetic architecture of complex traits and identify novel pleiotropic relationships. This PheWAS drew on five population-based studies representing four major racial/ethnic groups (European Americans (EA), African Americans (AA), Hispanics/Mexican-Americans, and Asian/Pacific Islanders) in PAGE, each site with measurements for multiple traits, associated laboratory measures, and intermediate biomarkers. A total of 83 single nucleotide polymorphisms (SNPs) identified by genome-wide association studies (GWAS) were genotyped across two or more PAGE study sites. Comprehensive tests of association, stratified by race/ethnicity, were performed, encompassing 4,706 phenotypes mapped to 105 phenotype-classes, and association results were compared across study sites. A total of 111 PheWAS results had significant associations for two or more PAGE study sites with consistent direction of effect with a significance threshold of p<0.01 for the same racial/ethnic group, SNP, and phenotype-class. Among results identified for SNPs previously associated with phenotypes such as lipid traits, type 2 diabetes, and body mass index, 52 replicated previously published genotype-phenotype associations, 26 represented phenotypes closely related to previously known genotype-phenotype associations, and 33 represented potentially novel genotype-phenotype associations with pleiotropic effects. The majority of the potentially novel results were for single PheWAS phenotype-classes, for example, for CDKN2A/B rs1333049 (previously associated with type 2 diabetes in EA) a PheWAS association was identified for hemoglobin levels in AA. Of note, however, GALNT2 rs2144300 (previously associated with high-density lipoprotein cholesterol levels in EA) had multiple potentially novel PheWAS associations, with hypertension related phenotypes in AA and with serum calcium levels and coronary artery disease phenotypes in EA. PheWAS identifies associations for hypothesis generation and exploration of the genetic architecture of complex traits.

Pleiotropic effects of genetic risk variants for other cancers on colorectal cancer risk: PAGE, GECCO and CCFR consortia.

OBJECTIVE: Genome-wide association studies have identified a large number of single nucleotide polymorphisms (SNPs) associated with a wide array of cancer sites. Several of these variants demonstrate associations with multiple cancers, suggesting pleiotropic effects and shared biological mechanisms across some cancers. We hypothesised that SNPs previously associated with other cancers may additionally be associated with colorectal cancer. In a large-scale study, we examined 171 SNPs previously associated with 18 different cancers for their associations with colorectal cancer. DESIGN: We examined 13 338 colorectal cancer cases and 40 967 controls from three consortia: Population Architecture using Genomics and Epidemiology (PAGE), Genetic Epidemiology of Colorectal Cancer (GECCO), and the Colon Cancer Family Registry (CCFR). Study-specific logistic regression results, adjusted for age, sex, principal components of genetic ancestry, and/or study specific factors (as relevant) were combined using fixed-effect meta-analyses to evaluate the association between each SNP and colorectal cancer risk. A Bonferroni-corrected p value of 2.92×10(-4) was used to determine statistical significance of the associations. RESULTS: Two correlated SNPs--rs10090154 and rs4242382--in Region 1 of chromosome 8q24, a prostate cancer susceptibility region, demonstrated statistically significant associations with colorectal cancer risk. The most significant association was observed with rs4242382 (meta-analysis OR=1.12; 95% CI 1.07 to 1.18; p=1.74×10(-5)), which also demonstrated similar associations across racial/ethnic populations and anatomical sub-sites. CONCLUSIONS: This is the first study to clearly demonstrate Region 1 of chromosome 8q24 as a susceptibility locus for colorectal cancer; thus, adding colorectal cancer to the list of cancer sites linked to this particular multicancer risk region at 8q24.

q-Diffusion leverages the full dimensionality of gene coexpression in single-cell transcriptomics.

Unlocking the full dimensionality of single-cell RNA sequencing data (scRNAseq) is the next frontier to a richer, fuller understanding of cell biology. We introduce q-diffusion, a framework for capturing the coexpression structure of an entire library of genes, improving on state-of-the-art analysis tools. The method is demonstrated via three case studies. In the first, q-diffusion helps gain statistical significance for differential effects on patient outcomes when analyzing the CALGB/SWOG 80405 randomized phase III clinical trial, suggesting precision guidance for the treatment of metastatic colorectal cancer. Secondly, q-diffusion is benchmarked against existing scRNAseq classification methods using an in vitro PBMC dataset, in which the proposed method discriminates IFN-γ stimulation more accurately. The same case study demonstrates improvements in unsupervised cell clustering with the recent Tabula Sapiens human atlas. Finally, a local distributional segmentation approach for spatial scRNAseq, driven by q-diffusion, yields interpretable structures of human cortical tissue.

A federated learning architecture for secure and private neuroimaging analysis.

The amount of biomedical data continues to grow rapidly. However, collecting data from multiple sites for joint analysis remains challenging due to security, privacy, and regulatory concerns. To overcome this challenge, we use federated learning, which enables distributed training of neural network models over multiple data sources without sharing data. Each site trains the neural network over its private data for some time and then shares the neural network parameters (i.e., weights and/or gradients) with a federation controller, which in turn aggregates the local models and sends the resulting community model back to each site, and the process repeats. Our federated learning architecture, MetisFL, provides strong security and privacy. First, sample data never leave a site. Second, neural network parameters are encrypted before transmission and the global neural model is computed under fully homomorphic encryption. Finally, we use information-theoretic methods to limit information leakage from the neural model to prevent a "curious" site from performing model inversion or membership attacks. We present a thorough evaluation of the performance of secure, private federated learning in neuroimaging tasks, including for predicting Alzheimer's disease and for brain age gap estimation (BrainAGE) from magnetic resonance imaging (MRI) studies in challenging, heterogeneous federated environments where sites have different amounts of data and statistical distributions.

Comparison of deep learning architectures for predicting amyloid positivity in Alzheimer's disease, mild cognitive impairment, and healthy aging, from T1-weighted brain structural MRI.

Abnormal ß-amyloid (Aß) accumulation in the brain is an early indicator of Alzheimer's disease (AD) and is typically assessed through invasive procedures such as PET (positron emission tomography) or CSF (cerebrospinal fluid) assays. As new anti-Alzheimer's treatments can now successfully target amyloid pathology, there is a growing interest in predicting Aß positivity (Aß+) from less invasive, more widely available types of brain scans, such as T1-weighted (T1w) MRI. Here we compare multiple approaches to infer Aß + from standard anatomical MRI: (1) classical machine learning algorithms, including logistic regression, XGBoost, and shallow artificial neural networks, (2) deep learning models based on 2D and 3D convolutional neural networks (CNNs), (3) a hybrid ANN-CNN, combining the strengths of shallow and deep neural networks, (4) transfer learning models based on CNNs, and (5) 3D Vision Transformers. All models were trained on paired MRI/PET data from 1,847 elderly participants (mean age: 75.1 yrs. ± 7.6SD; 863 females/984 males; 661 healthy controls, 889 with mild cognitive impairment (MCI), and 297 with Dementia), scanned as part of the Alzheimer's Disease Neuroimaging Initiative. We evaluated each model's balanced accuracy and F1 scores. While further tests on more diverse data are warranted, deep learning models trained on standard MRI showed promise for estimating Aß + status, at least in people with MCI. This may offer a potential screening option before resorting to more invasive procedures.

Comparison of Explainable AI Models for MRI-based Alzheimer's Disease Classification.

Deep learning models based on convolutional neural networks (CNNs) have been used to classify Alzheimer's disease or infer dementia severity from 3D T1-weighted brain MRI scans. Here, we examine the value of adding occlusion sensitivity analysis (OSA) and gradient-weighted class activation mapping (Grad-CAM) to these models to make the results more interpretable. Much research in this area focuses on specific datasets such as the Alzheimer's Disease Neuroimaging Initiative (ADNI) or National Alzheimer's Coordinating Center (NACC), which assess people of North American, predominantly European ancestry, so we examine how well models trained on these data generalize to a new population dataset from India (NIMHANS cohort). We also evaluate the benefit of using a combined dataset to train the CNN models. Our experiments show feature localization consistent with knowledge of AD from other methods. OSA and Grad-CAM resolve features at different scales to help interpret diagnostic inferences made by CNNs.

Expert-Level Detection of Referable Glaucoma from Fundus Photographs in a Safety Net Population: The AI and Teleophthalmology in Los Angeles Initiative.

Purpose: To develop and test a deep learning (DL) algorithm for detecting referable glaucoma in the Los Angeles County (LAC) Department of Health Services (DHS) teleretinal screening program. Methods: Fundus photographs and patient-level labels of referable glaucoma (defined as cup-to-disc ratio [CDR] ≥ 0.6) provided by 21 trained optometrist graders were obtained from the LAC DHS teleretinal screening program. A DL algorithm based on the VGG-19 architecture was trained using patient-level labels generalized to images from both eyes. Area under the receiver operating curve (AUC), sensitivity, and specificity were calculated to assess algorithm performance using an independent test set that was also graded by 13 clinicians with one to 15 years of experience. Algorithm performance was tested using reference labels provided by either LAC DHS optometrists or an expert panel of 3 glaucoma specialists. Results: 12,098 images from 5,616 patients (2,086 referable glaucoma, 3,530 non-glaucoma) were used to train the DL algorithm. In this dataset, mean age was 56.8 ± 10.5 years with 54.8% females and 68.2% Latinos, 8.9% Blacks, 2.7% Caucasians, and 6.0% Asians. 1,000 images from 500 patients (250 referable glaucoma, 250 non-glaucoma) with similar demographics (p ≥ 0.57) were used to test the DL algorithm. Algorithm performance matched or exceeded that of all independent clinician graders in detecting patient-level referable glaucoma based on LAC DHS optometrist (AUC = 0.92) or expert panel (AUC = 0.93) reference labels. Clinician grader sensitivity (range: 0.33-0.99) and specificity (range: 0.68-0.98) ranged widely and did not correlate with years of experience (p ≥ 0.49). Algorithm performance (AUC = 0.93) also matched or exceeded the sensitivity (range: 0.78-1.00) and specificity (range: 0.32-0.87) of 6 LAC DHS optometrists in the subsets of the test dataset they graded based on expert panel reference labels. Conclusions: A DL algorithm for detecting referable glaucoma developed using patient-level data provided by trained LAC DHS optometrists approximates or exceeds performance by ophthalmologists and optometrists, who exhibit variable sensitivity and specificity unrelated to experience level. Implementation of this algorithm in screening workflows could help reallocate eye care resources and provide more reproducible and timely glaucoma care.

Association of the FTO obesity risk variant rs8050136 with percentage of energy intake from fat in multiple racial/ethnic populations: the PAGE study.

Common obesity risk variants have been associated with macronutrient intake; however, these associations' generalizability across populations has not been demonstrated. We investigated the associations between 6 obesity risk variants in (or near) the NEGR1, TMEM18, BDNF, FTO, MC4R, and KCTD15 genes and macronutrient intake (carbohydrate, protein, ethanol, and fat) in 3 Population Architecture using Genomics and Epidemiology (PAGE) studies: the Multiethnic Cohort Study (1993-2006) (n = 19,529), the Atherosclerosis Risk in Communities Study (1987-1989) (n = 11,114), and the Epidemiologic Architecture for Genes Linked to Environment (EAGLE) Study, which accesses data from the Third National Health and Nutrition Examination Survey (1991-1994) (n = 6,347). We used linear regression, with adjustment for age, sex, and ethnicity, to estimate the associations between obesity risk genotypes and macronutrient intake. A fixed-effects meta-analysis model showed that the FTO rs8050136 A allele (n = 36,973) was positively associated with percentage of calories derived from fat (ßmeta = 0.2244 (standard error, 0.0548); P = 4 × 10(-5)) and inversely associated with percentage of calories derived from carbohydrate (ßmeta = -0.2796 (standard error, 0.0709); P = 8 × 10(-5)). In the Multiethnic Cohort Study, percentage of calories from fat assessed at baseline was a partial mediator of the rs8050136 effect on body mass index (weight (kg)/height (m)(2)) obtained at 10 years of follow-up (mediation of effect = 0.0823 kg/m(2), 95% confidence interval: 0.0559, 0.1128). Our data provide additional evidence that the association of FTO with obesity is partially mediated by dietary intake.

Post-genome-wide association study challenges for lipid traits: describing age as a modifier of gene-lipid associations in the Population Architecture using Genomics and Epidemiology (PAGE) study.

Numerous common genetic variants that influence plasma high-density lipoprotein cholesterol, low-density lipoprotein cholesterol (LDL-C), and triglyceride distributions have been identified via genome-wide association studies (GWAS). However, whether or not these associations are age-dependent has largely been overlooked. We conducted an association study and meta-analysis in more than 22,000 European Americans between 49 previously identified GWAS variants and the three lipid traits, stratified by age (males: <50 or ≥50 years of age; females: pre- or postmenopausal). For each variant, a test of heterogeneity was performed between the two age strata and significant Phet values were used as evidence of age-specific genetic effects. We identified seven associations in females and eight in males that displayed suggestive heterogeneity by age (Phet < 0.05). The association between rs174547 (FADS1) and LDL-C in males displayed the most evidence for heterogeneity between age groups (Phet = 1.74E-03, I(2) = 89.8), with a significant association in older males (P = 1.39E-06) but not younger males (P = 0.99). However, none of the suggestive modifying effects survived adjustment for multiple testing, highlighting the challenges of identifying modifiers of modest SNP-trait associations despite large sample sizes.

No evidence of interaction between known lipid-associated genetic variants and smoking in the multi-ethnic PAGE population.

Genome-wide association studies (GWAS) have identified many variants that influence high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, and/or triglycerides. However, environmental modifiers, such as smoking, of these known genotype-phenotype associations are just recently emerging in the literature. We have tested for interactions between smoking and 49 GWAS-identified variants in over 41,000 racially/ethnically diverse samples with lipid levels from the Population Architecture Using Genomics and Epidemiology (PAGE) study. Despite their biological plausibility, we were unable to detect significant SNP × smoking interactions.

Genetic variants associated with fasting glucose and insulin concentrations in an ethnically diverse population: results from the Population Architecture using Genomics and Epidemiology (PAGE) study.

BACKGROUND: Multiple genome-wide association studies (GWAS) within European populations have implicated common genetic variants associated with insulin and glucose concentrations. In contrast, few studies have been conducted within minority groups, which carry the highest burden of impaired glucose homeostasis and type 2 diabetes in the U.S. METHODS: As part of the 'Population Architecture using Genomics and Epidemiology (PAGE) Consortium, we investigated the association of up to 10 GWAS-identified single nucleotide polymorphisms (SNPs) in 8 genetic regions with glucose or insulin concentrations in up to 36,579 non-diabetic subjects including 23,323 European Americans (EA) and 7,526 African Americans (AA), 3,140 Hispanics, 1,779 American Indians (AI), and 811 Asians. We estimated the association between each SNP and fasting glucose or log-transformed fasting insulin, followed by meta-analysis to combine results across PAGE sites. RESULTS: Overall, our results show that 9/9 GWAS SNPs are associated with glucose in EA (p = 0.04 to 9 × 10-15), versus 3/9 in AA (p= 0.03 to 6 × 10-5), 3/4 SNPs in Hispanics, 2/4 SNPs in AI, and 1/2 SNPs in Asians. For insulin we observed a significant association with rs780094/GCKR in EA, Hispanics and AI only. CONCLUSIONS: Generalization of results across multiple racial/ethnic groups helps confirm the relevance of some of these loci for glucose and insulin metabolism. Lack of association in non-EA groups may be due to insufficient power, or to unique patterns of linkage disequilibrium.