Many roads to a gene-environment interaction.

Despite the importance of gene-environment interactions (GxEs) in improving and operationalizing genetic discovery, interpretation of any GxEs that are discovered can be surprisingly difficult. There are many potential biological and statistical explanations for a statistically significant finding and, likewise, it is not always clear what can be claimed based on a null result. A better understanding of the possible underlying mechanisms leading to a detected GxE can help investigators decide which are and which are not relevant to their hypothesis. Here, we provide a detailed explanation of five "phenomena," or data-generating mechanisms, that can lead to nonzero interaction estimates, as well as a discussion of specific instances in which they might be relevant. We hope that, given this framework, investigators can design more targeted experiments and provide cleaner interpretations of the associated results.

Multi-ancestry polygenic mechanisms of type 2 diabetes.

Type 2 diabetes (T2D) is a multifactorial disease with substantial genetic risk, for which the underlying biological mechanisms are not fully understood. In this study, we identified multi-ancestry T2D genetic clusters by analyzing genetic data from diverse populations in 37 published T2D genome-wide association studies representing more than 1.4 million individuals. We implemented soft clustering with 650 T2D-associated genetic variants and 110 T2D-related traits, capturing known and novel T2D clusters with distinct cardiometabolic trait associations across two independent biobanks representing diverse genetic ancestral populations (African, n = 21,906; Admixed American, n = 14,410; East Asian, n =2,422; European, n = 90,093; and South Asian, n = 1,262). The 12 genetic clusters were enriched for specific single-cell regulatory regions. Several of the polygenic scores derived from the clusters differed in distribution among ancestry groups, including a significantly higher proportion of lipodystrophy-related polygenic risk in East Asian ancestry. T2D risk was equivalent at a body mass index (BMI) of 30 kg m-2 in the European subpopulation and 24.2 (22.9-25.5) kg m-2 in the East Asian subpopulation; after adjusting for cluster-specific genetic risk, the equivalent BMI threshold increased to 28.5 (27.1-30.0) kg m-2 in the East Asian group. Thus, these multi-ancestry T2D genetic clusters encompass a broader range of biological mechanisms and provide preliminary insights to explain ancestry-associated differences in T2D risk profiles.

A Large-Scale Genome-Wide Study of Gene-Sleep Duration Interactions for Blood Pressure in 811,405 Individuals from Diverse Populations.

Although both short and long sleep duration are associated with elevated hypertension risk, our understanding of their interplay with biological pathways governing blood pressure remains limited. To address this, we carried out genome-wide cross-population gene-by-short-sleep and long-sleep duration interaction analyses for three blood pressure traits (systolic, diastolic, and pulse pressure) in 811,405 individuals from diverse population groups. We discover 22 novel gene-sleep duration interaction loci for blood pressure, mapped to genes involved in neurological, thyroidal, bone metabolism, and hematopoietic pathways. Non-overlap between short sleep (12) and long sleep (10) interactions underscores the plausibility of distinct influences of both sleep duration extremes in cardiovascular health. With several of our loci reflecting specificity towards population background or sex, our discovery sheds light on the importance of embracing granularity when addressing heterogeneity entangled in gene-environment interactions, and in therapeutic design approaches for blood pressure management.

Refinement of a published gene-physical activity interaction impacting HDL-cholesterol: role of sex and lipoprotein subfractions.

Large-scale gene-environment interaction (GxE) discovery efforts often involve compromises in the definition of outcomes and choice of covariates for the sake of data harmonization and statistical power. Consequently, refinement of exposures, covariates, outcomes, and population subsets may be helpful to establish often-elusive replication and evaluate potential clinical utility. Here, we used additional datasets, an expanded set of statistical models, and interrogation of lipoprotein metabolism via nuclear magnetic resonance (NMR)-based lipoprotein subfractions to refine a previously discovered GxE modifying the relationship between physical activity (PA) and HDL-cholesterol (HDL-C). This GxE was originally identified by Kilpeläinen et al., with the strongest cohort-specific signal coming from the Women's Genome Health Study (WGHS). We thus explored this GxE further in the WGHS (N = 23,294), with follow-up in the UK Biobank (UKB; N = 281,380), and the Multi-Ethnic Study of Atherosclerosis (MESA; N = 4,587). Self-reported PA (MET-hrs/wk), genotypes at rs295849 (nearest gene: LHX1), and NMR metabolomics data were available in all three cohorts. As originally reported, minor allele carriers of rs295849 in WGHS had a stronger positive association between PA and HDL-C (pint = 0.002). When testing a range of NMR metabolites (primarily lipoprotein and lipid subfractions) to refine the HDL-C outcome, we found a stronger interaction effect on medium-sized HDL particle concentrations (M-HDL-P; pint = 1.0×10-4) than HDL-C. Meta-regression revealed a systematically larger interaction effect in cohorts from the original meta-analysis with a greater fraction of women (p = 0.018). In the UKB, GxE effects were stronger both in women and using M-HDL-P as the outcome. In MESA, the primary interaction for HDL-C showed nominal significance (pint = 0.013), but without clear differences by sex and with a greater magnitude using large, rather than medium, HDL-P as an outcome. Towards reconciling these observations, further exploration leveraging NMR platform-specific HDL subfraction diameter annotations revealed modest agreement across all cohorts in the interaction affecting medium-to-large particles. Taken together, our work provides additional insights into a specific known gene-PA interaction while illustrating the importance of phenotype and model refinement towards understanding and replicating GxEs.

Re-analysis and meta-analysis of summary statistics from gene-environment interaction studies.

MOTIVATION: statistics from genome-wide association studies enable many valuable downstream analyses that are more efficient than individual-level data analysis while also reducing privacy concerns. As growing sample sizes enable better-powered analysis of gene-environment interactions, there is a need for gene-environment interaction-specific methods that manipulate and use summary statistics. RESULTS: We introduce two tools to facilitate such analysis, with a focus on statistical models containing multiple gene-exposure and/or gene-covariate interaction terms. REGEM (RE-analysis of GEM summary statistics) uses summary statistics from a single, multi-exposure genome-wide interaction study to derive analogous sets of summary statistics with arbitrary sets of exposures and interaction covariate adjustments. METAGEM (META-analysis of GEM summary statistics) extends current fixed-effects meta-analysis models to incorporate multiple exposures from multiple studies. We demonstrate the value and efficiency of these tools by exploring alternative methods of accounting for ancestry-related population stratification in genome-wide interaction study in the UK Biobank as well as by conducting a multi-exposure genome-wide interaction study meta-analysis in cohorts from the diabetes-focused ProDiGY consortium. These programs help to maximize the value of summary statistics from diverse and complex gene-environment interaction studies. AVAILABILITY AND IMPLEMENTATION: REGEM and METAGEM are open-source projects freely available at https://github.com/large-scale-gxe-methods/REGEM and https://github.com/large-scale-gxe-methods/METAGEM.

Type 2 Diabetes Modifies the Association of CAD Genomic Risk Variants With Subclinical Atherosclerosis.

BACKGROUND: Individuals with type 2 diabetes (T2D) have an increased risk of coronary artery disease (CAD), but questions remain about the underlying pathology. Identifying which CAD loci are modified by T2D in the development of subclinical atherosclerosis (coronary artery calcification [CAC], carotid intima-media thickness, or carotid plaque) may improve our understanding of the mechanisms leading to the increased CAD in T2D. METHODS: We compared the common and rare variant associations of known CAD loci from the literature on CAC, carotid intima-media thickness, and carotid plaque in up to 29 670 participants, including up to 24 157 normoglycemic controls and 5513 T2D cases leveraging whole-genome sequencing data from the Trans-Omics for Precision Medicine program. We included first-order T2D interaction terms in each model to determine whether CAD loci were modified by T2D. The genetic main and interaction effects were assessed using a joint test to determine whether a CAD variant, or gene-based rare variant set, was associated with the respective subclinical atherosclerosis measures and then further determined whether these loci had a significant interaction test. RESULTS: Using a Bonferroni-corrected significance threshold of P<1.6×10-4, we identified 3 genes (ATP1B1, ARVCF, and LIPG) associated with CAC and 2 genes (ABCG8 and EIF2B2) associated with carotid intima-media thickness and carotid plaque, respectively, through gene-based rare variant set analysis. Both ATP1B1 and ARVCF also had significantly different associations for CAC in T2D cases versus controls. No significant interaction tests were identified through the candidate single-variant analysis. CONCLUSIONS: These results highlight T2D as an important modifier of rare variant associations in CAD loci with CAC.

Multi-ancestry Polygenic Mechanisms of Type 2 Diabetes Elucidate Disease Processes and Clinical Heterogeneity.

We identified genetic subtypes of type 2 diabetes (T2D) by analyzing genetic data from diverse groups, including non-European populations. We implemented soft clustering with 650 T2D-associated genetic variants, capturing known and novel T2D subtypes with distinct cardiometabolic trait associations. The twelve genetic clusters were distinctively enriched for single-cell regulatory regions. Polygenic scores derived from the clusters differed in distribution between ancestry groups, including a significantly higher proportion of lipodystrophy-related polygenic risk in East Asian ancestry. T2D risk was equivalent at a BMI of 30 kg/m2 in the European subpopulation and 24.2 (22.9-25.5) kg/m2 in the East Asian subpopulation; after adjusting for cluster-specific genetic risk, the equivalent BMI threshold increased to 28.5 (27.1-30.0) kg/m2 in the East Asian group, explaining about 75% of the difference in BMI thresholds. Thus, these multi-ancestry T2D genetic subtypes encompass a broader range of biological mechanisms and help explain ancestry-associated differences in T2D risk profiles.

Multi-ancestry Polygenic Mechanisms of Type 2 Diabetes Elucidate Disease Processes and Clinical Heterogeneity.

We identified genetic subtypes of type 2 diabetes (T2D) by analyzing genetic data from diverse groups, including non-European populations. We implemented soft clustering with 650 T2D-associated genetic variants, capturing known and novel T2D subtypes with distinct cardiometabolic trait associations. The twelve genetic clusters were distinctively enriched for single-cell regulatory regions. Polygenic scores derived from the clusters differed in distribution between ancestry groups, including a significantly higher proportion of lipodystrophy-related polygenic risk in East Asian ancestry. T2D risk was equivalent at a BMI of 30 kg/m2 in the European subpopulation and 24.2 (22.9-25.5) kg/m2 in the East Asian subpopulation; after adjusting for cluster-specific genetic risk, the equivalent BMI threshold increased to 28.5 (27.1-30.0) kg/m2 in the East Asian group, explaining about 75% of the difference in BMI thresholds. Thus, these multi-ancestry T2D genetic subtypes encompass a broader range of biological mechanisms and help explain ancestry-associated differences in T2D risk profiles.

Clonal Hematopoiesis of Indeterminate Potential (CHIP) and Incident Type 2 Diabetes Risk.

OBJECTIVE: Clonal hematopoiesis of indeterminate potential (CHIP) is an aging-related accumulation of somatic mutations in hematopoietic stem cells, leading to clonal expansion. CHIP presence has been implicated in atherosclerotic coronary heart disease (CHD) and all-cause mortality, but its association with incident type 2 diabetes (T2D) is unknown. We hypothesized that CHIP is associated with elevated risk of T2D. RESEARCH DESIGN AND METHODS: CHIP was derived from whole-genome sequencing of blood DNA in the National Heart, Lung, and Blood Institute Trans-Omics for Precision Medicine (TOPMed) prospective cohorts. We performed analysis for 17,637 participants from six cohorts, without prior T2D, cardiovascular disease, or cancer. We evaluated baseline CHIP versus no CHIP prevalence with incident T2D, including associations with DNMT3A, TET2, ASXL1, JAK2, and TP53 variants. We estimated multivariable-adjusted hazard ratios (HRs) and 95% CIs with adjustment for age, sex, BMI, smoking, alcohol, education, self-reported race/ethnicity, and combined cohorts' estimates via fixed-effects meta-analysis. RESULTS: Mean (SD) age was 63.4 (11.5) years, 76% were female, and CHIP prevalence was 6.0% (n = 1,055) at baseline. T2D was diagnosed in n = 2,467 over mean follow-up of 9.8 years. Participants with CHIP had 23% (CI 1.04, 1.45) higher risk of T2D than those with no CHIP. Specifically, higher risk was for TET2 (HR 1.48; CI 1.05, 2.08) and ASXL1 (HR 1.76; CI 1.03, 2.99) mutations; DNMT3A was nonsignificant (HR 1.15; CI 0.93, 1.43). Statistical power was limited for JAK2 and TP53 analyses. CONCLUSIONS: CHIP was associated with higher incidence of T2D. CHIP mutations located on genes implicated in CHD and mortality were also related to T2D, suggesting shared aging-related pathology.

Association between timing and consistency of physical activity and type 2 diabetes: a cohort study on participants of the UK Biobank.

AIMS/HYPOTHESIS: We sought to quantify the relationship between morning, afternoon or evening physical activity and consistency (e.g. routine) and risk of type 2 diabetes. METHODS: A cohort of 93,095 UK Biobank participants (mean age 62 years) without a history of type 2 diabetes wore a wrist-worn accelerometer for 1 week. We converted accelerometer information to estimate metabolic equivalent of task (MET), summing MET h of total physical activity completed within three intra-day time segments (morning, afternoon and evening). We quantified physical activity consistency as the SD of participants' daily total physical activity. We ultimately associated each of the following with incident type 2 diabetes: (1) morning, afternoon or evening 'time-segmented' MET h per week; and (2) consistency. We also considered moderate-to-vigorous physical activity (MVPA) and vigorous physical activity (VPA) in association with type 2 diabetes incidence. RESULTS: When considering MET as the physical activity measure, we observed protective associations of morning (HR 0.90 [95% CI 0.86, 0.93], p=7×10-8) and afternoon (HR 0.91 [95% CI 0.87, 0.95], p=1×10-5) but did not have evidence for evening physical activity (HR 0.95 [95% CI 0.90, 1.00], p=0.07) with type 2 diabetes. There was no difference between MET-measured morning and afternoon physical activity. Our substitution model highlighted the importance of adjusting for lifestyle factors (e.g. sleep time and diet); the effect of a substitution between afternoon and evening physical activity was attenuated after adjustment for lifestyle variables. Consistency of MET-measured physical activity was not associated with type 2 diabetes (p=0.07). MVPA and VPA were associated with decreased risk for type 2 diabetes at all times of the day. CONCLUSIONS/INTERPRETATION: Total metabolic equivalents of physical activity in the morning and afternoon had a protective effect on diabetes risk and evening activity was not associated with diabetes. Consistency of physical activity did not play a role in decreasing risk for diabetes. Vigorous activity is associated with lower risk no matter the time of day of activity.

Corrigendum: Genetic heritability as a tool to evaluate the precision of 24-hour recall dietary questionnaire variables in UK Biobank.

[This corrects the article DOI: 10.3389/fgene.2022.1070511.].

Investigating Gene-Diet Interactions Impacting the Association Between Macronutrient Intake and Glycemic Traits.

Few studies have demonstrated reproducible gene-diet interactions (GDIs) impacting metabolic disease risk factors, likely due in part to measurement error in dietary intake estimation and insufficient capture of rare genetic variation. We aimed to identify GDIs across the genetic frequency spectrum impacting the macronutrient-glycemia relationship in genetically and culturally diverse cohorts. We analyzed 33,187 participants free of diabetes from 10 National Heart, Lung, and Blood Institute Trans-Omics for Precision Medicine program cohorts with whole-genome sequencing, self-reported diet, and glycemic trait data. We fit cohort-specific, multivariable-adjusted linear mixed models for the effect of diet, modeled as an isocaloric substitution of carbohydrate for fat, and its interactions with common and rare variants genome-wide. In main effect meta-analyses, participants consuming more carbohydrate had modestly lower glycemic trait values (e.g., for glycated hemoglobin [HbA1c], -0.013% HbA1c/250 kcal substitution). In GDI meta-analyses, a common African ancestry-enriched variant (rs79762542) reached study-wide significance and replicated in the UK Biobank cohort, indicating a negative carbohydrate-HbA1c association among major allele homozygotes only. Simulations revealed that >150,000 samples may be necessary to identify similar macronutrient GDIs under realistic assumptions about effect size and measurement error. These results generate hypotheses for further exploration of modifiable metabolic disease risk in additional cohorts with African ancestry. ARTICLE HIGHLIGHTS: We aimed to identify genetic modifiers of the dietary macronutrient-glycemia relationship using whole-genome sequence data from 10 Trans-Omics for Precision Medicine program cohorts. Substitution models indicated a modest reduction in glycemia associated with an increase in dietary carbohydrate at the expense of fat. Genome-wide interaction analysis identified one African ancestry-enriched variant near the FRAS1 gene that may interact with macronutrient intake to influence hemoglobin A1c. Simulation-based power calculations accounting for measurement error suggested that substantially larger sample sizes may be necessary to discover further gene-macronutrient interactions.

High-throughput genetic clustering of type 2 diabetes loci reveals heterogeneous mechanistic pathways of metabolic disease.

AIMS/HYPOTHESIS: Type 2 diabetes is highly polygenic and influenced by multiple biological pathways. Rapid expansion in the number of type 2 diabetes loci can be leveraged to identify such pathways. METHODS: We developed a high-throughput pipeline to enable clustering of type 2 diabetes loci based on variant-trait associations. Our pipeline extracted summary statistics from genome-wide association studies (GWAS) for type 2 diabetes and related traits to generate a matrix of 323 variants × 64 trait associations and applied Bayesian non-negative matrix factorisation (bNMF) to identify genetic components of type 2 diabetes. Epigenomic enrichment analysis was performed in 28 cell types and single pancreatic cells. We generated cluster-specific polygenic scores and performed regression analysis in an independent cohort (N=25,419) to assess for clinical relevance. RESULTS: We identified ten clusters of genetic loci, recapturing the five from our prior analysis as well as novel clusters related to beta cell dysfunction, pronounced insulin secretion, and levels of alkaline phosphatase, lipoprotein A and sex hormone-binding globulin. Four clusters related to mechanisms of insulin deficiency, five to insulin resistance and one had an unclear mechanism. The clusters displayed tissue-specific epigenomic enrichment, notably with the two beta cell clusters differentially enriched in functional and stressed pancreatic beta cell states. Additionally, cluster-specific polygenic scores were differentially associated with patient clinical characteristics and outcomes. The pipeline was applied to coronary artery disease and chronic kidney disease, identifying multiple overlapping clusters with type 2 diabetes. CONCLUSIONS/INTERPRETATION: Our approach stratifies type 2 diabetes loci into physiologically interpretable genetic clusters associated with distinct tissues and clinical outcomes. The pipeline allows for efficient updating as additional GWAS become available and can be readily applied to other conditions, facilitating clinical translation of GWAS findings. Software to perform this clustering pipeline is freely available.

Metabolomic markers of glucose regulation after a lifestyle intervention in prediabetes.

INTRODUCTION: Disentangling the specific factors that regulate glycemia from prediabetes to normoglycemia could improve type 2 diabetes prevention strategies. Metabolomics provides substantial insights into the biological understanding of environmental factors such as diet. This study aimed to identify metabolomic markers of regression to normoglycemia in the context of a lifestyle intervention (LSI) in individuals with prediabetes. RESEARCH DESIGN AND METHODS: We conducted a single-arm intervention study with 24 weeks of follow-up. Eligible study participants had at least one prediabetes criteria according to the American Diabetes Association guidelines, and body mass index between 25 and 45 kg/m2. LSI refers to a hypocaloric diet and >150 min of physical activity per week. Regression to normoglycemia (RNGR) was defined as achieving hemoglobin A1c (HbA1c) <5.5% in the final visit. Baseline and postintervention plasma metabolomic profiles were measured using liquid chromatography-tandem mass spectrometry. To select metabolites associated with RNGR, we conducted the least absolute shrinkage and selection operator-penalized regressions. RESULTS: The final sample was composed of 82 study participants. Changes in three metabolites were significantly associated with regression to normoglycemia; N-acetyl-D-galactosamine (OR=0.54; 95% CI 0.32 to 0.82), putrescine (OR=0.90, 95% CI 0.81 to 0.98), and 7-methylguanine (OR=1.06; 95% CI 1.02 to 1.17), independent of HbA1c and weight loss. In addition, metabolomic perturbations due to LSI displayed enrichment of taurine and hypotaurine metabolism pathway (p=0.03) compatible with biomarkers of protein consumption, lower red meat and animal fats and higher seafood and vegetables. CONCLUSIONS: Evidence from this study suggests that specific metabolomic markers have an influence on glucose regulation in individuals with prediabetes after 24 weeks of LSI independently of other treatment effects such as weight loss.

Whole genome sequence association analysis of fasting glucose and fasting insulin levels in diverse cohorts from the NHLBI TOPMed program.

The genetic determinants of fasting glucose (FG) and fasting insulin (FI) have been studied mostly through genome arrays, resulting in over 100 associated variants. We extended this work with high-coverage whole genome sequencing analyses from fifteen cohorts in NHLBI's Trans-Omics for Precision Medicine (TOPMed) program. Over 23,000 non-diabetic individuals from five race-ethnicities/populations (African, Asian, European, Hispanic and Samoan) were included. Eight variants were significantly associated with FG or FI across previously identified regions MTNR1B, G6PC2, GCK, GCKR and FOXA2. We additionally characterize suggestive associations with FG or FI near previously identified SLC30A8, TCF7L2, and ADCY5 regions as well as APOB, PTPRT, and ROBO1. Functional annotation resources including the Diabetes Epigenome Atlas were compiled for each signal (chromatin states, annotation principal components, and others) to elucidate variant-to-function hypotheses. We provide a catalog of nucleotide-resolution genomic variation spanning intergenic and intronic regions creating a foundation for future sequencing-based investigations of glycemic traits.

Variance-quantitative trait loci enable systematic discovery of gene-environment interactions for cardiometabolic serum biomarkers.

Gene-environment interactions represent the modification of genetic effects by environmental exposures and are critical for understanding disease and informing personalized medicine. These often induce differential phenotypic variance across genotypes; these variance-quantitative trait loci can be prioritized in a two-stage interaction detection strategy to greatly reduce the computational and statistical burden and enable testing of a broader range of exposures. We perform genome-wide variance-quantitative trait locus analysis for 20 serum cardiometabolic biomarkers by multi-ancestry meta-analysis of 350,016 unrelated participants in the UK Biobank, identifying 182 independent locus-biomarker pairs (p < 4.5×10-9). Most are concentrated in a small subset (4%) of loci with genome-wide significant main effects, and 44% replicate (p < 0.05) in the Women's Genome Health Study (N = 23,294). Next, we test each locus-biomarker pair for interaction across 2380 exposures, identifying 847 significant interactions (p < 2.4×10-7), of which 132 are independent (p < 0.05) after accounting for correlation between exposures. Specific examples demonstrate interaction of triglyceride-associated variants with distinct body mass- versus body fat-related exposures as well as genotype-specific associations between alcohol consumption and liver stress at the ADH1B gene. Our catalog of variance-quantitative trait loci and gene-environment interactions is publicly available in an online portal.

STAAR workflow: a cloud-based workflow for scalable and reproducible rare variant analysis.

SUMMARY: We developed the variant-Set Test for Association using Annotation infoRmation (STAAR) workflow description language (WDL) workflow to facilitate the analysis of rare variants in whole genome sequencing association studies. The open-access STAAR workflow written in the WDL allows a user to perform rare variant testing for both gene-centric and genetic region approaches, enabling genome-wide, candidate and conditional analyses. It incorporates functional annotations into the workflow as introduced in the STAAR method in order to boost the rare variant analysis power. This tool was specifically developed and optimized to be implemented on cloud-based platforms such as BioData Catalyst Powered by Terra. It provides easy-to-use functionality for rare variant analysis that can be incorporated into an exhaustive whole genome sequencing analysis pipeline. AVAILABILITY AND IMPLEMENTATION: The workflow is freely available from https://dockstore.org/workflows/github.com/sheilagaynor/STAAR_workflow. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Genetic heritability as a tool to evaluate the precision of 24-hour recall dietary questionnaire variables in UK Biobank.

A variety of statistical approaches in nutritional epidemiology have been developed to enhance the precision of dietary variables derived from longitudinal questionnaires. Correlation with biomarkers is often used to assess the relative validity of these different approaches, however, validated biomarkers do not always exist and are costly and laborious to collect. We present a novel high-throughput approach which utilizes the modest but importantly non-zero influence of genetic variation on variation in dietary intake to compare different statistical transformations of dietary variables. Specifically, we compare the heritability of crude averages with Empirical Bayes weighted averages for 302 correlated dietary variables from multiple 24-hour recall questionnaires in 177 K individuals in UK Biobank. Overall, the crude averages for frequency of consumption are more heritable than their Empirical Bayes counterparts only when the reliability of that item across questionnaires is high (measured by intra-class correlation), otherwise, the Empirical Bayes approach (for both unreliably measured frequencies and for average quantities independent of reliability) leads to higher heritability estimates. We also find that the more heritable versions of each dietary variable lead to stronger underlying statistical associations with specific genetic loci, many of which have well-known mechanisms, further supporting heritability as an alternative metric for relative validity in nutritional epidemiology and beyond.

GEM: scalable and flexible gene-environment interaction analysis in millions of samples.

MOTIVATION: Gene-environment interaction (GEI) studies are a general framework that can be used to identify genetic variants that modify the effects of environmental, physiological, lifestyle or treatment effects on complex traits. Moreover, accounting for GEIs can enhance our understanding of the genetic architecture of complex diseases and traits. However, commonly used statistical software programs for GEI studies are either not applicable to testing certain types of GEI hypotheses or have not been optimized for use in large samples. RESULTS: Here, we develop a new software program, GEM (Gene-Environment interaction analysis in Millions of samples), which supports the inclusion of multiple GEI terms, adjustment for GEI covariates and robust inference, while allowing multi-threading to reduce computation time. GEM can conduct GEI tests as well as joint tests of genetic main and interaction effects for both continuous and binary phenotypes. Through simulations, we demonstrate that GEM scales to millions of samples while addressing limitations of existing software programs. We additionally conduct a gene-sex interaction analysis on waist-hip ratio in 352 768 unrelated individuals from the UK Biobank, identifying 24 novel loci in the joint test that have not previously been reported in combined or sex-specific analyses. Our results demonstrate that GEM can facilitate the next generation of large-scale GEI studies and help advance our understanding of the genetic architecture of complex diseases and traits. AVAILABILITY AND IMPLEMENTATION: GEM is freely available as an open source project at https://github.com/large-scale-gxe-methods/GEM. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.