Efficient Bayesian mixed-model analysis increases association power in large cohorts

Application of Human Genetics to Prioritize Coagulation Cascade Protein Targets for Ischemic Stroke Prevention

BACKGROUND

While interindividual variations in concentration and function of coagulation cascade proteins are established risk factors for venous thromboembolism (VTE), their associations with arterial ischemic stroke are less well defined.

METHODS

We identified and validated genetic proxies for lifelong, randomized perturbations of coagulation cascade proteins in genome-wide association studies of circulating protein levels (deCODE, n=35 559; UK Biobank, n=46 218) and of VTE risk (81 190 cases and 1 419 671 controls). Study participants were all of European ancestry. We performed 2-sample Mendelian randomization and colocalization analyses to test associations of these genetic proxies with risk of ischemic stroke (62 100 cases and 1 234 808 controls from the GIGASTROKE consortium) and ischemic stroke subtypes, and further contextualized associations with VTE and secondary efficacy and safety outcomes.

RESULTS

We identified genetic proxies for 30 coagulation factors, with cross-trait associations recapitulating canonical coagulation biology. Mendelian randomization and colocalization analyses supported causal associations of genetically proxied levels of 5 proteins with risk of ischemic stroke, with all proteins associating with the cardioembolic stroke subtype: factor XI (odds ratio [OR] of cardioembolic stroke per 1-SD increase, 1.31 [95% CI, 1.19-1.44]; P=3.30×10-8), high-molecular-weight kininogen (OR, 1.19 [95% CI, 1.09-1.30]; P=7.79×10-5), prothrombin (OR, 1.83 [95% CI, 1.31-2.57]; P=4.20×10-4), soluble PROCR (protein C receptor; OR, 0.88 [95% CI, 0.82-0.95]; P=6.19×10-4), and γ' fibrinogen (OR per doubling in VTE risk due to lower γ' fibrinogen levels, 1.44 [95% CI, 1.25-1.66]; P=3.96×10-7). γ' Fibrinogen and prothrombin also associated with large artery atherosclerotic stroke, and no proteins were associated with small vessel stroke risk. By contrast, genetic proxies for several coagulation factors (including proteins C and S and factors V and VII) showed selective associations with VTE.

CONCLUSIONS

These data highlight specific coagulation cascade components implicated in ischemic stroke pathogenesis, while identifying proteins with distinct roles in VTE. These findings may inform development of novel anticoagulants and optimize their use in targeted populations with stroke.

Genome-wide association study identifying novel risk variants associated with glycaemic traits in the continental African AWI-Gen cohort

AIMS/HYPOTHESIS

Glycaemic traits such as high fasting glucose levels and insulin resistance are positively associated with the risk of type 2 diabetes and other cardiometabolic diseases. Genetic association studies have identified hundreds of associations for each glycaemic trait, yet very few studies have involved continental African populations. We report the results of genome-wide association studies (GWASs) in a pan-African cohort for four glycaemic traits, namely fasting glucose, fasting insulin, insulin resistance (HOMA-IR) and beta cell function (HOMA-B), which are quantitative variables that affect the risk of developing type 2 diabetes.

METHODS

GWASs for the four traits were conducted in approximately 10,000 individuals from the Africa Wits-INDEPTH Partnership for Genomics Studies (AWI-Gen) cohort, with participants from Burkina Faso, Ghana, Kenya and South Africa. Association testing was performed using linear mixed models implemented in BOLT-LMM, with age, sex, BMI and principal components as covariates. Replication, fine mapping and functional annotation were performed using standard approaches.

RESULTS

We identified a novel signal (rs574173815) in the intron of the ankyrin repeat domain 33B (ANKRD33B) gene associated with fasting glucose, and a novel signal (rs114029796) in the intronic region of the WD repeat domain 7 (WDR7) gene associated with fasting insulin. SNPs in WDR7 have been shown to be associated with type 2 diabetes. A variant (rs74806991) in the intron of ADAM metallopeptidase with thrombospondin type 1 motif 16 (ADAMTS16) and another variant (rs6506934) in the β-1,4-galactosyltransferase 6 gene (B4GALT6) are associated with HOMA-IR. Both ADAMTS16 and B4GALT6 are implicated in the development of type 2 diabetes. In addition, our study replicated several well-established fasting glucose signals in the GCK-YTK6, SLC2A2 and THORLNC gene regions.

CONCLUSIONS/INTERPRETATION

Our findings highlight the importance of performing GWASs for glycaemic traits in under-represented populations, especially continental African populations, to discover novel associated variants and broaden our knowledge of the genetic aetiology of glycaemic traits. The limited replication of well-known signals in this study hints at the possibility of a unique genetic architecture of these traits in African populations.

DATA AVAILABILITY

The dataset used in this study is available in the European Genome-Phenome Archive (EGA) database ( https://ega-archive.org/ ) under study accession code EGAS00001002482. The phenotype dataset accession code is EGAD00001006425 and the genotype dataset accession code is EGAD00010001996. The availability of these datasets is subject to controlled access by the Data and Biospecimen Access Committee of the H3Africa Consortium. GWAS summary statistics are accessible through the NHGRI-EBI GWAS Catalog ( https://www.ebi.ac.uk/gwas/ ).

Improved detection of microbiome-disease associations via population structure-aware generalized linear mixed effects models (microSLAM)

Microbiome association studies typically link host disease or other traits to summary statistics measured in metagenomics data, such as diversity or taxonomic composition. But identifying disease-associated species based on their relative abundance does not provide insight into why these microbes act as disease markers, and it overlooks cases where disease risk is related to specific strains with unique biological functions. To bridge this knowledge gap, we developed microSLAM, a mixed-effects model and an R package that performs association tests that connect host traits to the presence/absence of genes within each microbiome species, while accounting for strain genetic relatedness across hosts. Traits can be quantitative or binary (such as case/control). MicroSLAM is fit in three steps for each species. The first step estimates population structure across hosts. Step two calculates the association between population structure and the trait, enabling detection of species for which a subset of related strains confer risk. To identify specific genes whose presence/absence across diverse strains is associated with the trait, step three models the trait as a function of gene occurrence plus random effects estimated from step two. Applying microSLAM to 710 gut metagenomes from inflammatory bowel disease (IBD) samples, we discovered 56 species whose population structure correlates with IBD, meaning that different lineages are found in cases versus controls. After controlling for population structure, 20 species had genes significantly associated with IBD. Twenty-one of these genes were more common in IBD patients, while 32 genes were enriched in healthy controls, including a seven-gene operon in Faecalibacterium prausnitzii that is involved in utilization of fructoselysine from the gut environment. The vast majority of species detected by microSLAM were not significantly associated with IBD using standard relative abundance tests. These findings highlight the importance of accounting for within-species genetic variation in microbiome studies.

Bi-allelic variants in POPDC2 cause an autosomal recessive syndrome presenting with cardiac conduction defects and hypertrophic cardiomyopathy

POPDC2 encodes the Popeye domain-containing protein 2, which has an important role in cardiac pacemaking and conduction, due in part to its cyclic AMP (cAMP)-dependent binding and regulation of TREK-1 potassium channels. Loss of Popdc2 in mice results in sinus pauses and bradycardia, and morpholino-mediated knockdown of popdc2 in zebrafish results in atrioventricular (AV) block. We identified bi-allelic variants in POPDC2 in four families with a phenotypic spectrum consisting of sinus node dysfunction, AV conduction defects, and hypertrophic cardiomyopathy. Using homology modeling, we show that the identified variants are predicted to diminish the ability of POPDC2 to bind cAMP. In in vitro electrophysiological studies, we demonstrated that, in contrast with wild-type POPDC2, variants found in affected individuals failed to increase TREK-1 current density. While muscle biopsy of an affected individual did not show clear myopathic disease, it showed significantly reduced abundance of both POPDC1 and POPDC2, suggesting that stability and/or membrane trafficking of the POPDC1-POPDC2 complex is impaired by pathogenic variants in either protein. Single-cell RNA sequencing from human hearts demonstrated that co-expression of POPDC1 and POPDC2 was most prevalent in AV node, AV node pacemaker, and AV bundle cells. Using population-level genetic data of more than 1 million individuals, we show that none of the familial variants were associated with clinical outcomes in heterozygous state, suggesting that heterozygous family members are unlikely to develop clinical manifestations and therefore might not necessitate clinical follow-up. Our findings provide evidence for bi-allelic variants in POPDC2 causing a Mendelian autosomal recessive cardiac syndrome.

Improved genetic discovery and fine-mapping resolution through multivariate latent factor analysis of high-dimensional traits

Genome-wide association studies (GWASs) of high-dimensional traits, such as blood cell or metabolic traits, often use univariate approaches, ignoring trait relationships. Biological mechanisms generating variation in high-dimensional traits can be captured parsimoniously through a GWAS of latent factors. Here, we introduce flashfmZero, a zero-correlation latent-factor-based multi-trait fine-mapping approach. In an application to 25 latent factors derived from 99 blood cell traits in the INTERVAL cohort, we show that latent factor GWASs enable the detection of signals generating sub-threshold associations with several blood cell traits. The 99% credible sets (CS99) from flashfmZero were equal to or smaller in size than those from univariate fine-mapping of blood cell traits in 87% of our comparisons. In all cases univariate latent factor CS99 contained those from flashfmZero. Our latent factor approaches can be applied to GWAS summary statistics and will enhance power for the discovery and fine-mapping of associations for many traits.

Dissecting immune-mediated pathways in rheumatoid arthritis: A multivariate mediation analysis of antibodies and circulating proteins

Rheumatoid arthritis (RA) is a chronic inflammatory disorder with complex etiologies involving immune responses and circulating proteins. This study investigates the causal relationships between antibody immune responses, plasma circulating proteins, and the development of RA using Mendelian Randomization (MR) analysis; A two-sample and multivariate MR analysis was conducted to explore the mediating causal relationship between 46 antibody immune responses and RA through 4,907 plasma circulating proteins. Genetic variations were utilized as instrumental variables (IVs) to infer causality, ensuring that they met the assumptions of relevance, independence, and exclusion restriction. Data were sourced from the FinnGen R10 dataset, UK Biobank, and the SomaScan platform, providing a robust foundation for the analysis. Statistical methods including IVW, weighted median, and mode-based approaches were employed, complemented by sensitivity analyses to ensure the robustness of the findings; The study identified significant causal relationships between six antibody immune responses and RA, with three specific responses-Epstein-Barr virus EBNA-1, Epstein-Barr virus ZEBRA, and Anti-polyomavirus 2 IgG seropositivity-showing strong associations. However, reverse causality was detected for EBNA-1 and ZEBRA, leading to their exclusion from further analysis. Additionally, 12 plasma circulating proteins were found to have significant causal relationships with RA, with KCNIP3 emerging as a key protective factor. Multivariate MR analysis revealed that KCNIP3 mediates the relationship between Anti-polyomavirus 2 IgG seropositivity and RA, suggesting a potential protective mechanism. This study highlights the intricate relationships between specific antibody responses, circulating proteins, and RA risk. The findings suggest that certain proteins, particularly KCNIP3, may mediate the effects of immune responses on RA development, offering potential targets for therapeutic intervention.

Protein-truncating variants in UQCRC1 are associated with Parkinson's disease: evidence from half-million people

Recent studies have suggested a potential but inconsistent link between UQCRC1 and Parkinson's disease (PD). For the first time, we systematically investigated the association between non-synonymous variants in UQCRC1 and PD risk using data from the UK Biobank with half-million participants, which provide evidence supporting the role of UQCRC1 Protein-truncating variants (PTVs) in PD (P = 1.20 × 10-6, OR = 6.59) and highlight the importance of large-scale population studies in identifying rare genetic risk factors.

Genetic architecture and risk prediction of gestational diabetes mellitus in Chinese pregnancies

Gestational diabetes mellitus, a heritable metabolic disorder and the most common pregnancy-related condition, remains understudied regarding its genetic architecture and its potential for early prediction using genetic data. Here we conducted genome-wide association studies on 116,144 Chinese pregnancies, leveraging their non-invasive prenatal test sequencing data and detailed prenatal records. We identified 13 novel loci for gestational diabetes mellitus and 111 for five glycemic traits, with minor allele frequencies of 0.01-0.5 and absolute effect sizes of 0.03-0.62. Approximately 50% of these loci were specific to gestational diabetes mellitus and gestational glycemic levels, distinct from type 2 diabetes and general glycemic levels in East Asians. A machine learning model integrating polygenic risk scores and prenatal records predicted gestational diabetes mellitus before 20 weeks of gestation, achieving an area under the receiver operating characteristic curve of 0.729 and an accuracy of 0.835. Shapley values highlighted polygenic risk scores as key contributors. This model offers a cost-effective strategy for early gestational diabetes mellitus prediction using clinical non-invasive prenatal test.

Causal Relevance of Lp(a) for Coronary Heart Disease and Stroke Types in East Asian and European Ancestry Populations: A Mendelian Randomization Study

BACKGROUND

Elevated plasma levels of Lp(a) [lipoprotein(a)] are a causal risk factor for coronary heart disease and stroke in European individuals, but the causal relevance of Lp(a) for different stroke types and in East Asian individuals with different Lp(a) genetic architecture is uncertain.

METHODS

We measured plasma levels of Lp(a) in a nested case-control study of 18 174 adults (mean [SD] age, 57 [10] years; 49% female) in the China Kadoorie Biobank (CKB) and performed a genome-wide association analysis to identify genetic variants affecting Lp(a) levels, with replication in ancestry-specific subsets in UK Biobank. We further performed 2-sample Mendelian randomization analyses, associating ancestry-specific Lp(a)-associated instrumental variants derived from CKB or from published data in European individuals with risk of myocardial infarction (n=17 091), ischemic stroke (IS [n=29 233]) and its subtypes, or intracerebral hemorrhage (n=5845) in East Asian and European individuals using available data from CKB and genome-wide association analysis consortia.

RESULTS

In CKB observational analyses, plasma levels of Lp(a) were log-linearly and positively associated with higher risks of myocardial infarction and IS, but not with ICH. In genome-wide association analysis, we identified 29 single nucleotide polymorphisms independently associated with Lp(a) that together explained 33% of variance in Lp(a) in Chinese individuals. In UK Biobank, the lead Chinese variants identified in CKB were replicated in 1260 Chinese individuals, but explained only 10% of variance in Lp(a) in European individuals. In Mendelian randomization analyses, however, there were highly concordant effects of Lp(a) across both ancestries for all cardiovascular disease outcomes examined. In combined analyses of both ancestries, the proportional reductions in risk per 100 nmol/L lower genetically predicted Lp(a) levels for myocardial infarction were 3-fold greater than for total IS (rate ratio, 0.78 [95% CI, 0.76-0.81] versus 0.94 [0.92-0.96]), but were similar to those for large-artery IS (0.80 [0.73-0.87]; n=8134). There were weaker associations with cardioembolic IS (0.92 [95% CI, 0.86-0.98]; n=11 730), and no association with small-vessel IS (0.99 [0.91-1.07]; n=12 343) or with intracerebral hemorrhage (1.08 [0.96-1.21]; n=5845).

CONCLUSIONS

The effects of Lp(a) on risk of myocardial infarction and large-artery IS were comparable in East Asian and European individuals, suggesting that people with either ancestry could expect comparable proportional benefits for equivalent reductions in Lp(a), but there was little effect on other stroke types.

Assessing the causal effects of type 2 diabetes and obesity-related traits on COVID-19 severity

BACKGROUND

Type 2 diabetes (T2D) and obesity-related traits are highly comorbid with coronavirus disease 2019 (COVID-19), but their causal relationships with disease severity remain unclear. While recent Mendelian randomization (MR) studies suggest a causal link between obesity-related traits and COVID-19 severity, findings regarding T2D are inconsistent, particularly when adjusting for body mass index (BMI). This study aims to clarify these relationships.

METHODS

We applied various MR methods to assess the causal effects of BMI-adjusted T2D (T2DadjBMI) and obesity-related traits (BMI, waist circumference, and waist-hip ratio) on COVID-19 severity. Genetic instruments were obtained from large-scale genome-wide association studies (GWAS), including 898K participants for T2D and 2M for COVID-19 severity. To address potential bias from sample overlap, we conducted large-scale simulations comparing MR results from overlapping and independent samples.

RESULTS

Our MR analysis identified a significant causal relationship between T2DadjBMI and increased COVID-19 severity (OR = 1.057, 95% CI = 1.012-1.105). Obesity-related traits were also causally associated with COVID-19 severity. Simulations confirmed that MR results remained robust to sample overlap, demonstrating consistency between overlapping and independent datasets.

CONCLUSIONS

These findings highlight the causal role of T2D and obesity-related traits in COVID-19 severity, emphasizing the need for targeted prevention and management strategies for high-risk populations. The robustness of our MR analysis, even in the presence of sample overlap, strengthens the reliability of these causal inferences.

The phenotypic and genetic association between endometriosis and immunological diseases

STUDY QUESTION

Is there an increased risk of immunological diseases among endometriosis patients, and does a shared genetic basis contribute to this risk?

SUMMARY ANSWER

Endometriosis patients show a significantly increased risk of autoimmune, autoinflammatory, and mixed-pattern diseases, including rheumatoid arthritis, multiple sclerosis, coeliac disease, osteoarthritis, and psoriasis, with genetic correlations between endometriosis and osteoarthritis, rheumatoid arthritis, and multiple sclerosis, and a potential causal link to rheumatoid arthritis.

WHAT IS KNOWN ALREADY

The epidemiological evidence for an increased risk of immunological diseases among women with endometriosis is limited in scope and has varied in robustness due to the opportunity for biases. The presence of a biological basis for increased comorbidity across immunological conditions has not been investigated. Here we investigate the phenotypic and genetic association between endometriosis and 31 immune conditions in the UK Biobank.

STUDY DESIGN, SIZE, DURATION

Phenotypic analyses between endometriosis and immune conditions (17 classical autoimmune, 10 autoinflammatory, and 4 mixed-pattern diseases) were conducted using two approaches (8223 endometriosis, 64 620 immunological disease cases): (i) retrospective cohort study design to incorporate temporality between diagnoses and (ii) cross-sectional analysis for simple association. Genome-wide association studies (GWAS) and meta-analyses for those immune conditions that showed phenotypic association with endometriosis (1493-77 052 cases) were conducted.

PARTICIPANTS/MATERIALS, SETTING, METHODS

Comprehensive phenotypic association analyses were conducted in females in the UK Biobank. GWAS for immunological conditions were conducted in females-only and sex-combined study populations in UK Biobank and meta-analysed with existing largest available GWAS results. Genetic correlation and Mendelian randomization (MR) analyses were conducted to investigate potential causal relationships. Those immune conditions with significant genetic correlation with endometriosis were included in multi-trait analysis of GWAS to boost discovery of novel and shared genetic variants. These shared variants were functionally annotated to identify affected genes utilizing expression quantitative trait loci (eQTL) data from GTEx and eQTLGen databases. Biological pathway enrichment analysis was conducted to identify shared underlying biological pathways.

MAIN RESULTS AND THE ROLE OF CHANCE

In both retrospective cohort and cross-sectional analyses, endometriosis patients were at significantly increased (30-80%) risk of classical autoimmune (rheumatoid arthritis, multiple sclerosis, coeliac disease), autoinflammatory (osteoarthritis), and mixed-pattern (psoriasis) diseases. Osteoarthritis (genetic correlation (rg) = 0.28, P = 3.25 × 10-15), rheumatoid arthritis (rg = 0.27, P = 1.5 × 10-5) and multiple sclerosis (rg = 0.09, P = 4.00 × 10-3) were significantly genetically correlated with endometriosis. MR analysis suggested a causal association between endometriosis and rheumatoid arthritis (OR = 1.16, 95% CI = 1.02-1.33). eQTL analyses highlighted genes affected by shared risk variants, enriched for seven pathways across all four conditions, with three genetic loci shared between endometriosis and osteoarthritis (BMPR2/2q33.1, BSN/3p21.31, MLLT10/10p12.31) and one with rheumatoid arthritis (XKR6/8p23.1).

LIMITATIONS, REASONS FOR CAUTION

We conducted the first female-specific GWAS analyses for immune conditions. Given the novelty of these analyses, the sample sizes from which results were derived were limited compared to sex-combined GWAS meta-analyses, which limited the power to use female-specific summary statistics to uncover the shared genetic basis with endometriosis in follow-up analyses. Secondly, the 39 genome-wide significant endometriosis-associated variants used as instrumental variables in the MR analysis explained approximately 5% of disease variation, which may account for the nominal or non-significant MR results.

WIDER IMPLICATIONS OF THE FINDINGS

Endometriosis patients have a moderately increased risk for osteoarthritis, rheumatoid arthritis, and to a lesser extent, multiple sclerosis, due to underlying shared biological mechanisms. Clinical implications primarily involve the need for increased awareness and vigilance. The shared genetic basis opens up opportunities for developing new treatments or repurposing therapies across these conditions.

STUDY FUNDING/COMPETING INTEREST(S)

We thank all the UK Biobank and 23andMe participants. Part of this research was conducted using the UK Biobank Resource under Application Number 9637. N.R. was supported by a grant from the Wellbeing of Women UK (RG2031) and the EU Horizon 2020 funded project FEMaLe (101017562). A.P.M. was supported in part by Versus Arthritis (grant 21754). H.F. was supported by the National Natural Science Foundation of China (grant 32170663). N.R., S.A.M., and K.T.Z. were supported in part by a grant from CDMRP DoD PRMRP (W81XWH-20-PRMRP-IIRA). K.T.Z. and C.M.B. reported grants in 3 years prior, outside the submitted work, from Bayer AG, AbbVie Inc., Volition Rx, MDNA Life Sciences, PrecisionLife Ltd., and Roche Diagnostics Inc. S.A.M. reports grants in the 3 years prior, outside this submitted work, from AbbVie Inc. N.R. is a consultant for Endogene.bio, outside this submitted work. The other authors have no conflicts of interest to declare.

TRIAL REGISTRATION NUMBER

N/A.

GWAS Procedures for Gene Mapping in Diverse Populations With Complex Structures

With reduced genotyping costs, genome-wide association studies (GWAS) face more challenges in diverse populations with complex structures to map genes of interest. The complex structure demands sophisticated statistical models, and increased marker density and population size require efficient computing tools. Many statistical models and computing tools have been developed with varied properties in statistical power, computing efficiency, and user-friendly accessibility. Some statistical models were developed with dedicated computing tools, such as efficient mixed model analysis (EMMA), multiple loci mixed model (MLMM), fixed and random model circulating probability unification (FarmCPU), and Bayesian-information and linkage-disequilibrium iteratively nested keyway (BLINK). However, there are computing tools (e.g., GAPIT) that implement multiple statistical models, retain a constant user interface, and maintain enhancement on input data and result interpretation. In this study, we developed a protocol utilizing a minimal set of software tools (BEAGLE, BLINK, and GAPIT) to perform a variety of analyses including file format conversion, missing genotype imputation, GWAS, and interpretation of input data and outcome results. We demonstrated the protocol by reanalyzing data from the Rice 3000 Genomes Project and highlighting advancements in GWAS model development.

Divide and conquer approach for genome-wide association studies

Genome-wide association studies (GWAS) are computationally intensive, requiring significant time and resources with computational complexity scaling at least linearly with sample size. Here, we present an accurate and resource-efficient pipeline for GWAS that mitigates the impact of sample size on computational demands. Our approach involves (1) randomly partitioning the cohort into equally sized sub-cohorts, (2) conducting independent GWAS within each sub-cohort, and (3) integrating the results using a novel meta-analysis technique that accounts for population structure and other confounders between sub-cohorts. Importantly, we demonstrate through simulations and real-data examples in humans that our approach effectively manages analyzing related individuals, a critical factor in real datasets, while controlling for inflated effect sizes, a phenomenon known as winner's curse. We show that our method achieves the same discovery levels as standard approaches but with significantly reduced computational costs. Additionally, it is well-suited for incremental GWAS as new samples are added over time. Our implementation within a bioinformatics workflow management system enhances reproducibility and scalability.

Correcting for volunteer bias in GWAS increases SNP effect sizes and heritability estimates

Selection bias in genome-wide association studies (GWASs) due to volunteer-based sampling (volunteer bias) is poorly understood. The UK Biobank (UKB), one of the largest and most widely used cohorts, is highly selected. Using inverse probability (IP) weights we estimate inverse probability weighted GWAS (WGWAS) to correct GWAS summary statistics in the UKB for volunteer bias. Our IP weights were estimated using UK Census data - the largest source of population-representative data - made representative of the UKB's sampling population. These weights have a substantial SNP-based heritability of 4.8% (s.e. 0.8%), suggesting they capture volunteer bias in GWAS. Across ten phenotypes, WGWAS yields larger SNP effect sizes, larger heritability estimates, and altered gene-set tissue expression, despite decreasing the effective sample size by 62% on average, compared to GWAS. The impact of volunteer bias on GWAS results varies by phenotype. Traits related to disease, health behaviors, and socioeconomic status are most affected. We recommend that GWAS consortia provide population weights for their data sets, or use population-representative samples.

Gene clusters linked to insulin resistance identified in a genome-wide study of the Taiwan Biobank population

This pioneering genome-wide association study examined surrogate markers for insulin resistance (IR) in 147,880 Taiwanese individuals using data from the Taiwan Biobank. The study focused on two IR surrogate markers: the triglyceride to high-density lipoprotein cholesterol (TG:HDL-C) ratio and the TyG index (the product of fasting plasma glucose and triglycerides). We identified genome-wide significance loci within four gene clusters: GCKR, MLXIPL, APOA5, and APOC1, uncovering 197 genes associated with IR. Transcriptome-wide association analysis revealed significant associations between these clusters and TyG, primarily in adipose tissue. Gene ontology analysis highlighted pathways related to Alzheimer's disease, glucose homeostasis, insulin resistance, and lipoprotein dynamics. The study identified sex-specific genes associated with TyG. Polygenic risk score analysis linked both IR markers to gout and hyperlipidemia. Our findings elucidate the complex relationships between IR surrogate markers, genetic predisposition, and disease phenotypes in the Taiwanese population, contributing valuable insights to the field of metabolic research.

The genetic architecture of and evolutionary constraints on the human pelvic form

Human pelvic evolution following the human-chimpanzee divergence is thought to result in an obstetrical dilemma, a mismatch between large infant brains and narrowed female birth canals, but empirical evidence has been equivocal. By using deep learning on 31,115 dual-energy x-ray absorptiometry scans from UK Biobank, we identified 180 loci associated with seven highly heritable pelvic phenotypes. Birth canal phenotypes showed sex-specific genetic architecture, aligning with reproductive function. Larger birth canals were linked to slower walking pace and reduced back pain but increased hip osteoarthritis risk, whereas narrower birth canals were associated with reduced pelvic floor disorder risk but increased obstructed labor risk. Lastly, genetic correlation between birth canal and head widths provides evidence of coevolution between the human pelvis and brain, partially mitigating the dilemma.

A genotype-phenotype transformer to assess and explain polygenic risk

Genome-wide association studies have linked millions of genetic variants to biomedical phenotypes, but their utility has been limited by lack of mechanistic understanding and widespread epistatic interactions. Recently, Transformer models have emerged as a powerful machine learning architecture with potential to address these and other challenges. Accordingly, here we introduce the Genotype-to-Phenotype Transformer (G2PT), a framework for modeling hierarchical information flow among variants, genes, multigenic systems, and phenotypes. As proof-of-concept, we use G2PT to model the genetics of TG/HDL (triglycerides to high-density lipoprotein cholesterol), an indicator of metabolic health. G2PT predicts this trait via attention to 1,395 variants underlying at least 20 systems, including immune response and cholesterol transport, with accuracy exceeding state-of-the-art. It implicates 40 epistatic interactions, including epistasis between APOA4 and CETP in phospholipid transfer, a target pathway for cholesterol modification. This work positions hierarchical graph transformers as a next-generation approach to polygenic risk.

The genetics of low and high birthweight and their relationship with cardiometabolic disease

AIMS/HYPOTHESIS

Low birthweight infants are at increased risk not only of mortality, but also of type 2 diabetes mellitus and CVD in later life. At the opposite end of the spectrum, high birthweight infants have increased risk of birth complications, such as shoulder dystocia, neonatal hypoglycaemia and obesity, and similarly increased risk of type 2 diabetes mellitus and CVD. However, previous genome-wide association studies (GWAS) of birthweight in the UK Biobank have primarily focused on individuals within the 'normal' range and have excluded individuals with high and low birthweight (<2.5 kg or >4.5 kg). The aim of this study was to investigate genetic variation associated within the tail ends of the birthweight distribution, to: (1) see whether the genetic factors operating in these regions were different from those that explained variation in birthweight within the normal range; (2) explore the genetic correlation between extremes of birthweight and cardiometabolic disease; and (3) investigate whether analysing the full distribution of birthweight values, including the extremes, improved the ability to detect genuine loci in GWAS.

METHODS

We performed case-control GWAS analysis of low (<2.5 kg) and high (>4.5 kg) birthweight in the UK Biobank using REGENIE software (Nlow=20,947; Nhigh=12,715; Ncontrols=207,506) and conducted three continuous GWAS of birthweight, one including the full range of birthweights, one involving a truncated GWAS including only individuals with birthweights between 2.5 and 4.5 kg and a third GWAS that winsorised birthweight values <2.5 kg and >4.5 kg. Additionally, we performed bivariate linkage disequilibrium (LD) score regression to estimate the genetic correlation between low/normal/high birthweight and cardiometabolic traits.

RESULTS

Bivariate LD score regression analyses suggested that high birthweight had a mostly similar genetic aetiology to birthweight within the normal range (genetic correlation coefficient [rG]=0.91, 95% CI 0.83, 0.99), whereas there was more evidence for a separate set of genes underlying low birthweight (rG=-0.74, 95% CI 0.66, 0.82). Low birthweight was also significantly positively genetically correlated with most cardiometabolic traits and diseases we examined, whereas high birthweight was mostly positively genetically correlated with adiposity and anthropometric-related traits. The winsorisation strategy performed best in terms of locus detection, with the number of independent genome-wide significant associations (p<5×10-8) increasing from 120 genetic variants at 94 loci in the truncated GWAS to 270 genetic variants at 178 loci, including 27 variants at 25 loci that had not been identified in previous birthweight GWAS. This included a novel low-frequency missense variant in the ABCC8 gene, a gene known to be involved in congenital hyperinsulinism, neonatal diabetes mellitus and MODY, that was estimated to be responsible for a 170 g increase in birthweight amongst carriers.

CONCLUSIONS/INTERPRETATION

Our results underscore the importance of genetic factors in the genesis of the phenotypic correlation between birthweight and cardiometabolic traits and diseases.

Pitfalls in performing genome-wide association studies on ratio traits

Genome-wide association studies (GWASs) are often performed on ratios composed of a numerator trait divided by a denominator trait. Examples include body mass index (BMI) and the waist-to-hip ratio, among many others. Explicitly or implicitly, the goal of forming the ratio is typically to adjust for an association between the numerator and denominator. While forming ratios may be clinically expedient, there are several important issues with performing GWAS on ratios. Forming a ratio does not "adjust" for the denominator in the sense of conditioning on it, and it is unclear whether associations with ratios are attributable to the numerator, the denominator, or both. Here we demonstrate that associations arising in ratio GWAS can be entirely denominator driven, implying that at least some associations uncovered by ratio GWAS may be due solely to a putative adjustment variable. In a survey of 10 common ratio traits, we find that the ratio model disagrees with the adjusted model (performing GWAS on the numerator while conditioning on the denominator) at around 1/3 of loci. Using BMI as an example, we show that variants detected by only the ratio model are more strongly associated with the denominator (height), while variants detected by only the adjusted model are more strongly associated with the numerator (weight). Although the adjusted model provides effect sizes with a clearer interpretation, it is susceptible to collider bias. We propose and validate a simple method of correcting for the genetic component of collider bias via leave-one-chromosome-out polygenic scoring.

Cross-ancestry analyses of Chinese and European populations reveal insights into the genetic architecture and disease implication of metabolites

Differential susceptibilities to various diseases and corresponding metabolite variations have been documented across diverse ethnic populations, but the genetic determinants of these disparities remain unclear. Here, we performed large-scale genome-wide association studies of 171 directly quantifiable metabolites from a nuclear magnetic resonance-based metabolomics platform in 10,792 Han Chinese individuals. We identified 15 variant-metabolite associations, eight of which were successfully replicated in an independent Chinese population (n = 4,480). By cross-ancestry meta-analysis integrating 213,397 European individuals from the UK Biobank, we identified 228 additional variant-metabolite associations and improved fine-mapping precision. Moreover, two-sample Mendelian randomization analyses revealed evidence that genetically predicted levels of triglycerides in high-density lipoprotein were associated with a higher risk of coronary artery disease and that of glycine with a lower risk of heart failure in both ancestries. These findings enhance our understanding of the shared and specific genetic architecture of metabolites as well as their roles in complex diseases across populations.

Association of MTHFR missense variants with thromboembolic diseases and coagulation factor levels in European populations

BACKGROUND

Investigations of the association between missense variants in the methylenetetrahydrofolate reductase (MTHFR) gene and thromboembolic diseases have been limited by small sample sizes. The effect of these variants on coagulation factor levels remains similarly uncertain.

OBJECTIVES

To test the association of the C677T and A1298C missense variants in MTHFR with risk of venous thromboembolism (VTE), cardioembolic stroke (CES), and circulating coagulation cascade protein levels.

PATIENTS/METHODS

We analyzed genetic associations of MTHFR missense variants with VTE (81,190 cases and 1,419,671 controls), CES (10,804 cases and 1,234,808 controls), and circulating levels of coagulation cascade proteins from the deCODE (n = 35,559) and UK Biobank (n = 46,218) cohorts. All participants in these genetic analyses were of European ancestry. We report odds ratios (OR) and beta coefficients per copy of the missense variant. VTE associations were compared to the effect of the Factor V Leiden variant.

RESULTS

The A1298C variant conferred a small increased risk of VTE (OR per allele: 1.03, 95% confidence interval [CI] 1.02-1.04, P = 1.36 × 10- 6). This effect was 30-fold weaker than the effect of Factor V Leiden on VTE. After correction for multiple comparisons, the C677T variant did not demonstrate a significant association with VTE (OR 0.99, 95% CI 0.98-1.00, P = 0.04). Neither variant was associated with CES (P ≥ 0.18), nor with any of the 34 coagulation cascade proteins after correction for multiple comparisons.

CONCLUSIONS

These data do not support a role for MTHFR genetic testing as part of an inherited thrombophilia evaluation.

Genomic dissection of sleep archetypes in a large autism cohort

Poor sleep is a major concern among individuals with autism and their caregivers. To better characterize the genetic and phenotypic heterogeneity of poor sleep in autism, we recruited 5,686 families from SPARK, a nationwide genetic study of autism, who described their sleep experiences using the Children's Sleep Health Questionnaire (CSHQ) and other self-report items. The collective experiences from this large sample allowed us to discover eight distinct archetypes of sleep in autism. Membership in some of these archetypes showed significant SNP-heritability (0.50 - 0.65, 95% confidence interval = 0.08 - 1), and polygenic estimates of educational attainment, BMI, and ADHD risk contributed extensively to the genetic signatures of these sleep archetypes. Surprisingly, polygenic estimates of general population sleep phenotypes showed sparser and more modest associations, perhaps suggesting that the genetic drivers of disordered sleep in autism may be distinct from those encountered in the general population. GWAS on archetype membership yielded no genome-wide significant loci, however, the most significant gene for the most severe archetype was the nitric oxide (NO) signaling gene NOS1AP, which was previously linked to sleep disruption in schizophrenia. Finally, the eight sleep archetypes showed specific signatures of treatment response across five major categories of sleep aid, pointing to the potential of treatment plans that are tailored to the nature of the sleep problem. These findings provide critical new insight into the comorbidities, subtypes, and genetic risk factors associated with disordered sleep in autism.

Secure and federated genome-wide association studies for biobank-scale datasets

Sharing data across institutions for genome-wide association studies (GWAS) would enhance the discovery of genetic variation linked to health and disease1,2. However, existing data-sharing regulations limit the scope of such collaborations3. Although cryptographic tools for secure computation promise to enable collaborative analysis with formal privacy guarantees, existing approaches either are computationally impractical or do not implement current state-of-the-art methods4-6. We introduce secure federated genome-wide association studies (SF-GWAS), a combination of secure computation frameworks and distributed algorithms that empowers efficient and accurate GWAS on private data held by multiple entities while ensuring data confidentiality. SF-GWAS supports widely used GWAS pipelines based on principal-component analysis or linear mixed models. We demonstrate the accuracy and practical runtimes of SF-GWAS on five datasets, including a UK Biobank cohort of 410,000 individuals, showcasing an order-of-magnitude improvement in runtime compared to previous methods. Our work enables secure collaborative genomic studies at unprecedented scale.

Family-based genome-wide association study designs for increased power and robustness

Family-based genome-wide association studies (FGWASs) use random, within-family genetic variation to remove confounding from estimates of direct genetic effects (DGEs). Here we introduce a 'unified estimator' that includes individuals without genotyped relatives, unifying standard and FGWAS while increasing power for DGE estimation. We also introduce a 'robust estimator' that is not biased in structured and/or admixed populations. In an analysis of 19 phenotypes in the UK Biobank, the unified estimator in the White British subsample and the robust estimator (applied without ancestry restrictions) increased the effective sample size for DGEs by 46.9% to 106.5% and 10.3% to 21.0%, respectively, compared to using genetic differences between siblings. Polygenic predictors derived from the unified estimator demonstrated superior out-of-sample prediction ability compared to other family-based methods. We implemented the methods in the software package snipar in an efficient linear mixed model that accounts for sample relatedness and sibling shared environment.

Canine genome-wide association study identifies DENND1B as an obesity gene in dogs and humans

Obesity is a heritable disease, but its genetic basis is incompletely understood. Canine population history facilitates trait mapping. We performed a canine genome-wide association study for body condition score-a measure of obesity-in 241 Labrador retrievers. Using a cross-species approach, we showed that canine obesity genes are also associated with rare and common forms of obesity in humans. The lead canine association was within the gene DENN domain containing 1B (DENND1B). Each copy of the alternate allele was associated with ~7.5% greater body fat. We demonstrate a role for this gene in regulating signaling and trafficking of melanocortin 4 receptor, a critical controller of energy homeostasis. Thus, canine genetics identified obesity genes and mechanisms relevant to both dogs and humans.

Missense variants in FRS3 affect body mass index in populations of diverse ancestries

Obesity is associated with adverse effects on health and quality of life. Improved understanding of its underlying pathophysiology is essential for developing counteractive measures. To search for sequence variants with large effects on BMI, we perform a multi-ancestry meta-analysis of 13 genome-wide association studies on BMI, including data derived from 1,534,555 individuals of European ancestry, 339,657 of Asian ancestry, and 130,968 of African ancestry. We identify an intergenic 262,760 base pair deletion at the MC4R locus that associates with 4.11 kg/m2 higher BMI per allele, likely through downregulation of MC4R. Moreover, a rare FRS3 missense variant, p.Glu115Lys, only found in individuals from Finland, associates with 1.09 kg/m2 lower BMI per allele. We also detect three other low-frequency FRS3 missense variants that associate with BMI with smaller effects and are enriched in different ancestries. We characterize FRS3 as a BMI-associated gene, encoding an adaptor protein known to act downstream of BDNF and TrkB, which regulate appetite, food intake, and energy expenditure through unknown signaling pathways. The work presented here contributes to the biological foundation of obesity by providing a convincing downstream component of the BDNF-TrkB pathway, which could potentially be targeted for obesity treatment.

Effect of vitamin A on adult lung function: a triangulation of evidence approach

BACKGROUND

Vitamin A, an essential micronutrient obtained through the diet, plays a crucial role in lung development and contributes to lung regeneration. We aimed to investigate its effect on adult lung function using triangulation of evidence from both observational and genetic data.

METHODS

Using data on 150 000 individuals from UK Biobank and correcting for measurement error (generalised structural equation modelling), we first investigated the association of dietary vitamin A intake (total vitamin A, carotene and retinol) with lung function (forced vital capacity (FVC), forced expiratory volume in 1 s (FEV1)/FVC)). We then assessed the causality of these associations using Mendelian randomisation (MR), and we investigated the effects on adult lung function of 39 genes related to vitamin A, and their interaction with vitamin A intake.

FINDINGS

Our observational analysis suggests a positive association between carotene intake and FVC only (13.3 mL per 100 µg/day; p=2.9×10-9), with stronger associations in smokers, but no association of retinol intake with FVC or FEV1/FVC. The MR similarly shows a beneficial effect of serum beta-carotene on FVC only, with no effect of serum retinol on FVC nor FEV1/FVC. Nine of the vitamin A-related genes were associated with adult lung function, six of which have not been previously identified in genome-wide studies and three (NCOA2, RDH10, RXRB) in any type of genetic study of lung function. Five genes showed possible gene-vitamin A intake interactions.

INTERPRETATION

Our triangulation study provides convincing evidence for a causal effect of vitamin A, carotene in particular, on adult lung function, suggesting a beneficial effect of a carotene-rich diet on adult lung health.

Evaluation of the causal relationship between 28 circulating biomarkers and osteoarthritis : a bidirectional Mendelian randomization study

Aims

Circulating biochemistry markers are commonly used to monitor and detect disease-induced dysfunctions including osteoarthritis (OA). However, the causal nature of this relationship is nevertheless largely unknown, due to unmeasured confounding factors from observational studies. We aimed to reveal the causal relationship between 28 circulating biochemistry markers and OA pathogenesis.

Methods

We conducted a comprehensive bidirectional two-sample Mendelian randomization (MR) study between 28 circulating biomarkers and six OA types, using large-scale genome-wide association study (GWAS) summary statistics data from a UK Biobank cohort (n = 450,243) and the latest OA meta-analysis (n = 826,690). We replicated the significant results of low-density lipoprotein cholesterol (LDL-C) and total cholesterol (TC) in an independent large GWAS dataset obtained from the Global Lipids Genetics Consortium (GLGC) (n > 800,000).

Results

Using 73 to 792 instrumental variables for biomarkers, this large MR analysis identified 11 causal associations at the Bonferroni corrected significance level of 2.98 × 10-4, involving seven biomarkers and five OA types. LDL-C (odds ratio (OR) per SD increase 0.90, 95% CI 0.86 to 0.93), apolipoprotein B (OR 0.86, 95% CI 0.82 to 0.91), TC (OR 0.90, 95% CI 0.86 to 0.94), calcium (OR 0.82, 95% CI 0.75 to 0.90), and glucose (OR 0.81, 95% CI 0.73 to 0.89) are causally associated with a reduced risk of OA, while phosphate (OR 1.18, 95% CI 1.08 to 1.30) and aspartate aminotransferase (OR 1.15, 95% CI 1.07 to 1.24) are causally associated with an increased risk. Analysis of GLGC summary statistics successfully replicated LDL-C (OR 0.93, 95% CI 0.90 to 0.96) and TC (OR 0.92, 95% CI 0.89 to 0.95).

Conclusion

This comprehensive bidirectional MR analysis provides new insights into the prevention and treatment of OA, as well as understanding the biological mechanism underlying OA pathogenesis.

The integration of quantile regression with 3VmrMLM identifies more QTNs and QTN-by-environment interactions using SNP- and haplotype-based markers

Current methods used in genome-wide association studies frequently lack power owing to their inability to detect heterogeneous associations and rare and multiallelic variants. To address these issues, quantile regression is integrated with a three (compressed) variance component multi-locus random-SNP-effect mixed linear model (3VmrMLM) to propose q3VmrMLM for detecting heterogeneous quantitative trait nucleotides (QTNs) and QTN-by-environment interactions (QEIs), and then design haplotype-based q3VmrMLM (q3VmrMLM-Hap) for identifying multiallelic haplotypes and rare variants. In Monte Carlo simulation studies, q3VmrMLM had higher power than 3VmrMLM, sequence kernel association test (SKAT), and integrated quantile rank test (iQRAT). In a re-analysis of 10 traits in 1439 rice hybrids, 261 known genes were identified only by q3VmrMLM and q3VmrMLM-Hap, whereas 175 known genes were detected by both the new and existing methods. Of all the significant QTNs with known genes, q3VmrMLM (179: 140 variance heterogeneity and 157 quantile effect heterogeneity) found more heterogeneous QTNs than 3VmrMLM (123), SKAT (27), and iQRAT (29); q3VmrMLM-Hap (121) mapped more low-frequency (<0.05) QTNs than q3VmrMLM (51), 3VmrMLM (43), SKAT (11), and iQRAT (12); and q3VmrMLM-Hap (12), q3VmrMLM (16), and 3VmrMLM (12) had similar power in identifying gene-by-environment interactions. All significant and suggested QTNs achieved the highest predictive accuracy (r = 0.9045). In conclusion, this study describes a new and complementary approach to mining genes and unraveling the genetic architecture of complex traits in crops.

Dissecting genetic regulation of metabolic coordination

Understanding genetic regulation of metabolism is critical for gaining insights into the causes of metabolic diseases. Traditional metabolome-based genome-wide association studies (mGWAS) focus on static associations between single nucleotide polymorphisms (SNPs) and metabolite levels, overlooking the changing relationships caused by genotypes within the metabolic network. Notably, some metabolites exhibit changes in correlation patterns with other metabolites under certain physiological conditions while maintaining their overall abundance level. In this manuscript, we develop Metabolic Differential-coordination GWAS (mdGWAS), an innovative framework that detects SNPs associated with the changing correlation patterns between metabolites and metabolic pathways. This approach transcends and complements conventional mean-based analyses by identifying latent regulatory factors that govern the system-level metabolic coordination. Through comprehensive simulation studies, mdGWAS demonstrated robust performance in detecting SNP-metabolite-metabolite associations. Applying mdGWAS to genotyping and mass spectrometry (MS)-based metabolomics data of the METabolic Syndrome In Men (METSIM) Study revealed novel SNPs and genes potentially involved in the regulation of the coordination between metabolic pathways.

Evaluation of genomic selection models using whole genome sequence data and functional annotation in Belgian Blue cattle

BACKGROUND

The availability of large cohorts of whole-genome sequenced individuals, combined with functional annotation, is expected to provide opportunities to improve the accuracy of genomic selection (GS). However, such benefits have not often been observed in initial applications. The reference population for GS in Belgian Blue Cattle (BBC) continues to grow. Combined with the availability of reference panels of sequenced individuals, it provides an opportunity to evaluate GS models using whole genome sequence (WGS) data and functional annotation.

RESULTS

Here, we used data from 16,508 cows, with phenotypes for five muscular development traits and imputed at the WGS level, in combination with in silico functional annotation and catalogs of putative regulatory variants obtained from experimental data. We evaluated first GS models using the entire WGS data, with or without functional annotation. At this marker density, we were able to run two approaches, assuming either a highly polygenic architecture (GBLUP) or allowing some variants to have larger effects (BayesRR-RC, a Bayesian mixture model), and observed an increased reliability compared to the official GBLUP model at medium marker density (on average 0.016 and 0.018 for GBLUP and BayesRR-RC, respectively). When functional annotation was used, we observed slightly higher reliabilities with an extension of GBLUP that included multiple polygenic terms (one per functional group), while reliabilities decreased with BayesRR-RC. We then used large subsets of variants selected based on functional information or with a linkage disequilibrium (LD) pruning approach, which allowed us to evaluate two additional approaches, BayesCπ and Bayesian Sparse Linear Mixed Model (BSLMM). Reliabilities were higher for these panels than for the WGS data, with the highest accuracies obtained when markers were selected based on functional information. In our setting, BSLMM systematically achieved higher reliabilities than other methods.

CONCLUSIONS

GS with large panels of functional variants selected from WGS data allowed a significant increase in reliability compared to the official genomic evaluation approach. However, the benefits of using WGS and functional data remained modest, indicating that there is still room for improvement, for example by further refining the functional annotation in the BBC breed.

Unsupervised Learning-Derived Complex Metabolic Signatures Refine Cardiometabolic Risk

BACKGROUND

Cardiometabolic diseases have become a leading cause of morbidity and mortality globally. Nuclear magnetic resonance metabolomics represents a precise tool for assessing metabolic individuality.

OBJECTIVES

This study aimed to use unsupervised learning to decode plasma metabolomic profiles, providing new insights into the etiology of cardiometabolic diseases.

METHODS

We applied unsupervised learning to generate robust metabolic signatures from the plasma profiles of 118,001 UK Biobank participants. Phenome-wide and genome-wide association studies were conducted to reveal their phenomic and genetic architectures. Integrated prospective cohort analyses and Mendelian randomization clarified their roles in cardiometabolic risks.

RESULTS

Eleven metabolic clusters were identified, linked to 101 loci and 445 phenotypes, mostly cardiometabolic diseases. These novel signatures partially outperformed traditional lipids in cardiometabolic risk prediction. Triglyceride-rich lipoproteins demonstrated superior predictive power for ischemic heart disease, type 2 diabetes, and hypertension, compared with apolipoprotein B and lipoprotein(a). Non-high-density lipoprotein cholesterol was found to increase the risk of hyperlipidemia and ischemic heart disease while offering a protective effect against type 2 diabetes. Furthermore, different high-density lipoprotein clusters showed heterogeneous associations with cardiometabolic diseases, with high-density lipoprotein subpopulations enriched in free cholesterol or triglyceride increasing risk, and those enriched in cholesterol esters providing protection.

CONCLUSIONS

These metabolic signatures extract comprehensive information from the metabolomic profile while maintaining clarity and interpretability, facilitating clinical translation. The findings emphasize the crucial roles of lipid subpopulations in cardiometabolic risks, encouraging clinicians to take a more nuanced approach to managing blood lipids and balancing different disease risks.

Large-scale genome-wide association analyses identify novel genetic loci and mechanisms in hypertrophic cardiomyopathy

Hypertrophic cardiomyopathy (HCM) is an important cause of morbidity and mortality with both monogenic and polygenic components. Here, we report results from a large genome-wide association study and multitrait analysis including 5,900 HCM cases, 68,359 controls and 36,083 UK Biobank participants with cardiac magnetic resonance imaging. We identified 70 loci (50 novel) associated with HCM and 62 loci (20 novel) associated with relevant left ventricular traits. Among the prioritized genes in the HCM loci, we identify a novel HCM disease gene, SVIL, which encodes the actin-binding protein supervillin, showing that rare truncating SVIL variants confer a roughly tenfold increased risk of HCM. Mendelian randomization analyses support a causal role of increased left ventricular contractility in both obstructive and nonobstructive forms of HCM, suggesting common disease mechanisms and anticipating shared response to therapy. Taken together, these findings increase our understanding of the genetic basis of HCM, with potential implications for disease management.

A comprehensive spatio-cellular map of the human hypothalamus

The hypothalamus is a brain region that plays a key role in coordinating fundamental biological functions1. However, our understanding of the underlying cellular components and neurocircuitries have, until recently, emerged primarily from rodent studies2,3. Here we combine single-nucleus sequencing of 433,369 human hypothalamic cells with spatial transcriptomics, generating a comprehensive spatio-cellular transcriptional map of the hypothalamus, the 'HYPOMAP'. Although conservation of neuronal cell types between humans and mice, as based on transcriptomic identity, is generally high, there are notable exceptions. Specifically, there are significant disparities in the identity of pro-opiomelanocortin neurons and in the expression levels of G-protein-coupled receptors between the two species that carry direct implications for currently approved obesity treatments. Out of the 452 hypothalamic cell types, we find that 291 neuronal clusters are significantly enriched for expression of body mass index (BMI) genome-wide association study genes. This enrichment is driven by 426 'effector' genes. Rare deleterious variants in six of these (MC4R, PCSK1, POMC, CALCR, BSN and CORO1A) associate with BMI at population level, and CORO1A has not been linked previously to BMI. Thus, HYPOMAP provides a detailed atlas of the human hypothalamus in a spatial context and serves as an important resource to identify new druggable targets for treating a wide range of conditions, including reproductive, circadian and metabolic disorders.

Genetic subtyping of obesity reveals biological insights into the uncoupling of adiposity from its cardiometabolic comorbidities

Obesity is a highly heterogeneous disease that cannot be captured by one single adiposity trait. Here, we performed a multi-trait analysis to study obesity in the context of its common cardiometabolic comorbidities, acknowledging that not all individuals with obesity suffer from cardiometabolic comorbidities and that not all those with normal weight clinically present without them. We leveraged individual-level genotype-phenotype data of 452,768 individuals from the UK Biobank and designed uncoupling phenotypes that are continuous and range from high adiposity with a healthy cardiometabolic profile to low adiposity with an unhealthy cardiometabolic profile. Genome-wide association analyses of these uncoupling phenotypes identified 266 independent variants across 205 genomic loci where the adiposity-increasing allele is also associated with a lower cardiometabolic risk. Consistent with the individual variant effects, a genetic score (GRSuncoupling) that aggregates the uncoupling effects of the 266 variants was associated with lower risk of cardiometabolic disorders, including dyslipidemias (OR=0.92, P=1.4×10-89), type 2 diabetes (OR=0.94, P=6×10-21), and ischemic heart disease (OR=0.96, P=7×10-11), despite a higher risk of obesity (OR=1.16, P=4×10-108), which is in sharp contrast to the association profile observed for the adiposity score (GRSBFP). Nevertheless, a higher GRSuncoupling score was also associated with a higher risk of other, mostly weight-bearing disorders, to the same extent as the GRSBFP. The 266 variants clustered into eight subsets, each representing a genetic subtype of obesity with a distinct cardiometabolic risk profile, characterized by specific underlying pathways. Association of GRSuncoupling and GRSBFP with levels of 2,920 proteins in plasma found 208 proteins to be associated with both scores. The majority (85%) of these overlapping GRS-protein associations were directionally consistent, suggesting adiposity-driven effects. In contrast, levels of 32 (15%) proteins (e.g. IGFBP1, IGFBP2, LDLR, SHBG, MSTN) had opposite directional effects between GRSBFP and GRSuncoupling, suggesting that cardiometabolic health, and not adiposity, associated with their levels. Follow-up analyses provide further support for adipose tissue expandability, insulin secretion and beta-cell function, beiging of white adipose tissue, inflammation and fibrosis. They also highlight mechanisms not previously implicated in uncoupling, such as hepatic lipid accumulation, hepatic control of glucose homeostasis, and skeletal muscle growth and function. Taken together, our findings contribute new insights into the mechanisms that uncouple adiposity from its cardiometabolic comorbidities and illuminate some of the heterogeneity of obesity, which is critical for advancing precision medicine.

Rare genetic associations with human lifespan in UK Biobank are enriched for oncogenic genes

Human lifespan is shaped by genetic and environmental factors. To enable precision health, understanding how genetic variants influence mortality is essential. We conducted a survival analysis in European ancestry participants of the UK Biobank, using age-at-death (N=35,551) and last-known-age (N=358,282). The associations identified were predominantly driven by cancer. We found lifespan-associated loci (APOE, ZSCAN23) for common variants and six genes where burden of loss-of-function variants were linked to reduced lifespan (TET2, ATM, BRCA2, CKMT1B, BRCA1, ASXL1). Additionally, eight genes with pathogenic missense variants were associated with reduced lifespan (DNMT3A, SF3B1, TET2, PTEN, SOX21, TP53, SRSF2, RLIM). Many of these genes are involved in oncogenic pathways and clonal hematopoiesis. Our findings highlight the importance of understanding genetic factors driving the most prevalent causes of mortality at a population level, highlighting the potential of early genetic testing to identify germline and somatic variants increasing one's susceptibility to cancer and/or early death.

The times they are AI-changing: AI-powered advances in the application of extracellular vesicles to liquid biopsy in breast cancer

Artificial intelligence (AI) is revolutionizing scientific research by facilitating a paradigm shift in data analysis and discovery. This transformation is characterized by a fundamental change in scientific methods and concepts due to AI's ability to process vast datasets with unprecedented speed and accuracy. In breast cancer research, AI aids in early detection, prognosis, and personalized treatment strategies. Liquid biopsy, a noninvasive tool for detecting circulating tumor traits, could ideally benefit from AI's analytical capabilities, enhancing the detection of minimal residual disease and improving treatment monitoring. Extracellular vesicles (EVs), which are key elements in cell communication and cancer progression, could be analyzed with AI to identify disease-specific biomarkers. AI combined with EV analysis promises an enhancement in diagnosis precision, aiding in early detection and treatment monitoring. Studies show that AI can differentiate cancer types and predict drug efficacy, exemplifying its potential in personalized medicine. Overall, the integration of AI in biomedical research and clinical practice promises significant changes and advancements in diagnostics, personalized medicine-based approaches, and our understanding of complex diseases like cancer.

The impact of sleep on breast cancer-specific mortality: a Mendelian randomisation study

BACKGROUND

The relationship between sleep traits and survival in breast cancer is uncertain and complex. There are multiple biological, psychological and treatment-related factors that could link sleep and cancer outcomes. Previous studies could be biased due to methodological limitations such as reverse causation and confounding. Here, we used two-sample mendelian randomisation (MR) to investigate the causal relationship between sleep and breast cancer mortality.

METHODS

Publicly available genetic summary data from females of European ancestry from UK Biobank and 23andme and the Breast Cancer Association Consortium were used to generate instrumental variables for sleep traits (chronotype, insomnia symptoms, sleep duration, napping, daytime-sleepiness, and ease of getting up (N = 446,118-1,409,137)) and breast cancer outcomes (15 years post-diagnosis, stratified by tumour subtype and treatment (N = 91,686 and Ndeaths = 7,531 over a median follow-up of 8.1 years)). Sensitivity analyses were used to assess the robustness of analyses to MR assumptions.

RESULTS

Initial results found some evidence for a per category increase in daytime-sleepiness reducing overall breast cancer mortality (HR = 0.34, 95% CI = 0.14, 0.80), and for insomnia symptoms reducing odds of mortality in oestrogen receptor positive breast cancers not receiving chemotherapy (HR = 0.18, 95% CI = 0.05, 0.68) and in patients receiving aromatase inhibitors (HR = 0.23, 95% CI = 0.07, 0.78). Importantly, these relationships were not robust following sensitivity analyses meaning we could not demonstrate any causal relationships.

CONCLUSIONS

This study did not provide evidence that sleep traits have a causal role in breast cancer mortality. Further work characterising disruption to normal sleep behaviours and its effects on tumour biology, treatment compliance and quality of life are needed.

A variant in HMMR/HMMR-AS1 is associated with serum alanine aminotransferase levels in the Ryukyu population

The Ryukyu archipelago is located southwest of the Japanese islands, and people originally from this region, the Ryukyu population, have a unique genetic background distinct from that of other populations, including people from mainland Japan. However, few genetic studies have focused on the Ryukyu population. In this study, we performed genome-wide association studies (GWAS) on the serum levels of alanine aminotransferase (ALT, n = 15,224), aspartate aminotransferase (AST, n = 15,203), and gamma-glutamyl transferase (GGT, n = 14,496) in the Ryukyu population. We found 13 loci with a genome-wide significant association (P < 5 × 10-8), three for ALT, four for AST, and six for GGT, including one novel locus associated with ALT: rs117595134-A in HMMR/HMMR-AS1, ß =  - 0.131, standard error = 0.024, P = 4.90 × 10-8. Rs117595134-A is common in the Japanese population but is not observed in other ethnic populations in the 1000 genomes database. Additionally, 77 of 80 loci derived from Korean GWAS and 541 of 716 loci from European GWAS showed the same directions of effect (P = 1.41 × 10-19, P = 2.50 × 10-44, binomial test), indicating that most of susceptibility loci are shared between the Ryukyu population and other ethnic populations.

Comparative studies of 2168 plasma proteins measured by two affinity-based platforms in 4000 Chinese adults

Proteomics offers unique insights into human biology and drug development, but few studies have directly compared the utility of different proteomics platforms. We measured plasma levels of 2168 proteins in 3976 Chinese adults using both Olink Explore and SomaScan platforms. The correlation of protein levels between platforms was modest (median rho = 0.29), with protein abundance and data quality parameters being key factors influencing correlation. For 1694 proteins with one-to-one matched reagents, 765 Olink and 513 SomaScan proteins had cis-pQTLs, including 400 with colocalising cis-pQTLs. Moreover, 1096 Olink and 1429 SomaScan proteins were associated with BMI, while 279 and 154 proteins were associated with risk of ischaemic heart disease, respectively. Addition of Olink and SomaScan proteins to conventional risk factors for ischaemic heart disease improved C-statistics from 0.845 to 0.862 (NRI: 12.2%) and 0.863 (NRI: 16.4%), respectively. These results demonstrate the utility of these platforms and could inform the design and interpretation of future studies.

Instability of high polygenic risk classification and mitigation by integrative scoring

Polygenic risk scores (PRS) continue to improve with novel methods and expanding genome-wide association studies. Healthcare and commercial laboratories are increasingly deploying PRS reports to patients, but it is unknown how the classification of high polygenic risk changes across individual PRS. Here, we assess the association and classification performance of cataloged PRS for three complex traits. We chronologically order all trait-related publications (Pubn) and identify the single PRS Best(Pubn) for each Pubn that has the strongest association with the target outcome. While each Best(Pubn) demonstrates generally consistent population-level strengths of associations, the classification of individuals in the top 10% of each Best(Pubn) distribution varies widely. Using the PRSmix framework, which integrates information across several PRS to improve prediction, we generate corresponding ChronoAdd(Pubn) scores for each Pubn that combine all polygenic scores from all publications up to and including Pubn. When compared with Best(Pubn), ChronoAdd(Pubn) scores demonstrate more consistent high-risk classification amongst themselves. This integrative scoring approach provides stable and reliable classification of high-risk individuals and is an adaptable framework into which new scores can be incorporated as they are introduced, integrating easily with current PRS implementation strategies.

SPAGRM: effectively controlling for sample relatedness in large-scale genome-wide association studies of longitudinal traits

Sample relatedness is a major confounder in genome-wide association studies (GWAS), potentially leading to inflated type I error rates if not appropriately controlled. A common strategy is to incorporate a random effect related to genetic relatedness matrix (GRM) into regression models. However, this approach is challenging for large-scale GWAS of complex traits, such as longitudinal traits. Here we propose a scalable and accurate analysis framework, SPAGRM, which controls for sample relatedness via a precise approximation of the joint distribution of genotypes. SPAGRM can utilize GRM-free models and thus is applicable to various trait types and statistical methods, including linear mixed models and generalized estimation equations for longitudinal traits. A hybrid strategy incorporating saddlepoint approximation greatly increases the accuracy to analyze low-frequency and rare genetic variants, especially in unbalanced phenotypic distributions. We also introduce SPAGRM(CCT) to aggregate the results following different models via Cauchy combination test. Extensive simulations and real data analyses demonstrated that SPAGRM maintains well-controlled type I error rates and SPAGRM(CCT) can serve as a broadly effective method. Applying SPAGRM to 79 longitudinal traits extracted from UK Biobank primary care data, we identified 7,463 genetic loci, making a pioneering attempt to conduct GWAS for these traits as longitudinal traits.

Analysis-ready VCF at Biobank scale using Zarr

Background

Variant Call Format (VCF) is the standard file format for interchanging genetic variation data and associated quality control metrics. The usual row-wise encoding of the VCF data model (either as text or packed binary) emphasises efficient retrieval of all data for a given variant, but accessing data on a field or sample basis is inefficient. Biobank scale datasets currently available consist of hundreds of thousands of whole genomes and hundreds of terabytes of compressed VCF. Row-wise data storage is fundamentally unsuitable and a more scalable approach is needed.

Results

Zarr is a format for storing multi-dimensional data that is widely used across the sciences, and is ideally suited to massively parallel processing. We present the VCF Zarr specification, an encoding of the VCF data model using Zarr, along with fundamental software infrastructure for efficient and reliable conversion at scale. We show how this format is far more efficient than standard VCF based approaches, and competitive with specialised methods for storing genotype data in terms of compression ratios and single-threaded calculation performance. We present case studies on subsets of three large human datasets (Genomics England: n=78,195; Our Future Health: n=651,050; All of Us: n=245,394) along with whole genome datasets for Norway Spruce (n=1,063) and SARS-CoV-2 (n=4,484,157). We demonstrate the potential for VCF Zarr to enable a new generation of high-performance and cost-effective applications via illustrative examples using cloud computing and GPUs.

Conclusions

Large row-encoded VCF files are a major bottleneck for current research, and storing and processing these files incurs a substantial cost. The VCF Zarr specification, building on widely-used, open-source technologies has the potential to greatly reduce these costs, and may enable a diverse ecosystem of next-generation tools for analysing genetic variation data directly from cloud-based object stores, while maintaining compatibility with existing file-oriented workflows.

A Litmus Test for Confounding in Polygenic Scores

Polygenic scores (PGSs) are being rapidly adopted for trait prediction in the clinic and beyond. PGSs are often thought of as capturing the direct genetic effect of one's genotype on their phenotype. However, because PGSs are constructed from population-level associations, they are influenced by factors other than direct genetic effects, including stratification, assortative mating, and dynastic effects ("SAD effects"). Our interpretation and application of PGSs may hinge on the relative impact of SAD effects, since they may often be environmentally or culturally mediated. We developed a method that estimates the proportion of variance in a PGS (in a given sample) that is driven by direct effects, SAD effects, and their covariance. We leverage a comparison of a PGS of interest based on a standard GWAS with a PGS based on a sibling GWAS-which is largely immune to SAD effects-to quantify the relative contribution of each type of effect to variance in the PGS of interest. Our method, Partitioning Genetic Scores Using Siblings (PGSUS, pron. "Pegasus"), breaks down variance components further by axes of genetic ancestry, allowing for a nuanced interpretation of SAD effects. In particular, PGSUS can detect stratification along major axes of ancestry as well as SAD variance that is "isotropic" with respect to axes of ancestry. Applying PGSUS, we found evidence of stratification in PGSs constructed using large meta-analyses of height and educational attainment as well as in a range of PGSs constructed using the UK Biobank. In some instances, a given PGS appears to be stratified along a major axis of ancestry in one prediction sample but not in another (for example, in comparisons of prediction in samples from different countries, or in ancient DNA vs. contemporary samples). Finally, we show that different approaches for adjustment for population structure in GWASs have distinct advantages with respect to mitigation of ancestry-axis-specific and isotropic SAD variance in PGS. Our study illustrates how family-based designs can be combined with standard population-based designs to guide the interpretation and application of genomic predictors.

Sequence-based GWAS in 180,000 German Holstein cattle reveals new candidate variants for milk production traits

BACKGROUND

Milk production traits are complex and influenced by many genetic and environmental factors. Although extensive research has been performed for these traits, with many associations unveiled thus far, due to their crucial economic importance, complex genetic architecture, and the fact that causal variants in cattle are still scarce, there is a need for a better understanding of their genetic background. In this study, we aimed to identify new candidate loci associated with milk production traits in German Holstein cattle, the most important dairy breed in Germany and worldwide. For that purpose, 180,217 cattle were imputed to the sequence level and large-scale genome-wide association study (GWAS) followed by fine-mapping and evolutionary and functional annotation were carried out to identify and prioritize new association signals.

RESULTS

Using the imputed sequence data of a large cattle dataset, we identified 50,876 significant variants, confirming many known and identifying previously unreported candidate variants for milk (MY), fat (FY), and protein yield (PY). Genome-wide significant signals were fine-mapped with the Bayesian approach that determines the credible variant sets and generates the probability of causality for each signal. The variants with the highest probabilities of being causal were further classified using external information about the function and evolution, making the prioritization for subsequent validation experiments easier. The top potential causal variants determined with fine-mapping explained a large percentage of genetic variance compared to random ones; 178 variants explained 11.5%, 104 explained 7.7%, and 68 variants explained 3.9% of the variance for MY, FY, and PY, respectively, demonstrating the potential for causality.

CONCLUSIONS

Our findings proved the power of large samples and sequence-based GWAS in detecting new association signals. In order to fully exploit the power of GWAS, one should aim at very large samples combined with whole-genome sequence data. These can also come with both computational and time burdens, as presented in our study. Although milk production traits in cattle are comprehensively investigated, the genetic background of these traits is still not fully understood, with the potential for many new associations to be revealed, as shown. With constantly growing sample sizes, we expect more insights into the genetic architecture of milk production traits in the future.

A scalable variational inference approach for increased mixed-model association power

The rapid growth of modern biobanks is creating new opportunities for large-scale genome-wide association studies (GWASs) and the analysis of complex traits. However, performing GWASs on millions of samples often leads to trade-offs between computational efficiency and statistical power, reducing the benefits of large-scale data collection efforts. We developed Quickdraws, a method that increases association power in quantitative and binary traits without sacrificing computational efficiency, leveraging a spike-and-slab prior on variant effects, stochastic variational inference and graphics processing unit acceleration. We applied Quickdraws to 79 quantitative and 50 binary traits in 405,088 UK Biobank samples, identifying 4.97% and 3.25% more associations than REGENIE and 22.71% and 7.07% more than FastGWA. Quickdraws had costs comparable to REGENIE, FastGWA and SAIGE on the UK Biobank Research Analysis Platform service, while being substantially faster than BOLT-LMM. These results highlight the promise of leveraging machine learning techniques for scalable GWASs without sacrificing power or robustness.

Fine-scale population structure and widespread conservation of genetic effect sizes between human groups across traits

Understanding genetic differences between populations is essential for avoiding confounding in genome-wide association studies and improving polygenic score (PGS) portability. We developed a statistical pipeline to infer fine-scale Ancestry Components and applied it to UK Biobank data. Ancestry Components identify population structure not captured by widely used principal components, improving stratification correction for geographically correlated traits. To estimate the similarity of genetic effect sizes between groups, we developed ANCHOR, which estimates changes in the predictive power of an existing PGS in distinct local ancestry segments. ANCHOR infers highly similar (estimated correlation 0.98 ± 0.07) effect sizes between UK Biobank participants of African and European ancestry for 47 of 53 quantitative phenotypes, suggesting that gene-environment and gene-gene interactions do not play major roles in poor cross-ancestry PGS transferability for these traits in the United Kingdom, and providing optimism that shared causal mutations operate similarly in different populations.

Effects of childhood and adult height on later life cardiovascular disease risk estimated through Mendelian randomization

Taller individuals are at elevated and protected risk of various cardiovascular disease endpoints. Whether this is due to a direct consequence of their height during childhood, a long-term effect of remaining tall throughout the lifecourse, or confounding by other factors, is unknown. We sought to address this by harnessing human genetic data from the UK Biobank to separate the independent effects of childhood and adulthood height using an approach known as lifecourse Mendelian randomization (MR). Protective effects of taller childhood height on risk of later life coronary artery disease (OR = 0.78 per change in height category, 95% CI = 0.70 to 0.86, P = 4 × 10- 10) and stroke (OR = 0.93, 95% CI = 0.86 to 1.00, P = 0.03) using data from large-scale consortia were found using a univariable model, although evidence of these effects attenuated in a multivariable setting upon accounting for adulthood height. In contrast, direct effects of taller childhood height on increased risk of later life atrial fibrillation (OR = 1.61, 95% CI = 1.42 to 1.79, P = 5 × 10- 7) and thoracic aortic aneurysm (OR = 1.55, 95% CI = 1.16 to 1.95, P = 0.03) were found even after accounting for adulthood height. Evidence for both of these direct effects was replicated in the Million Veterans Program. The protective effect of childhood height on risk of coronary artery disease and stroke can be largely explained by taller children typically becoming taller individuals in later life. Conversely, the independent effect of childhood height on increased risk of atrial fibrillation and thoracic aortic aneurysm may point towards developmental mechanisms in early life which confer a lifelong risk on these disease outcomes.

Exploring the role of circulating proteins in multiple myeloma risk: a Mendelian randomization study

Multiple myeloma (MM) is an incurable blood cancer with unclear aetiology. Proteomics is a valuable tool in exploring mechanisms of disease. We investigated the causal relationship between circulating proteins and MM risk, using two of the largest cohorts with proteomics data to-date. We performed bidirectional two-sample Mendelian randomization (MR; forward MR = causal effect estimation of proteins and MM risk; reverse MR = causal effect estimation of MM risk and proteins). Summary statistics for plasma proteins were obtained from genome-wide association studies performed using SomaLogic (N = 35,559; deCODE) and Olink (N = 34,557; UK Biobank; UKB) proteomic platforms and for MM risk from a meta-analysis of UKB and FinnGen (case = 1649; control = 727,247) or FinnGen only (case = 1085; control = 271,463). Cis-SNPs associated with protein levels were used to instrument circulating proteins. We evaluated proteins for the consistency of directions of effect across MR analyses (with 95% confidence intervals not overlapping the null) and corroborating evidence from genetic colocalization. In the forward MR, 994 (SomaLogic) and 1570 (Olink) proteins were instrumentable. 440 proteins were analysed in both deCODE and UKB; 302 (69%) of these showed consistent directions of effect in the forward MR. Seven proteins had 95% confidence intervals (CIs) that did not overlap the null in both forward MR analyses and did not have evidence for an effect in the reverse direction: higher levels of dermatopontin (DPT), beta-crystallin B1 (CRYBB1), interleukin-18-binding protein (IL18BP) and vascular endothelial growth factor receptor 2 (KDR) and lower levels of odorant-binding protein 2b (OBP2B), glutamate-cysteine ligase regulatory subunit (GCLM) and gamma-crystallin D (CRYGD) were implicated in increasing MM risk. Evidence from genetic colocalization did not meet our threshold for a shared causal signal between any of these proteins and MM risk (h4 < 0.8). Our results highlight seven circulating proteins which may be involved in MM risk. Although evidence from genetic colocalization suggests these associations may not be robust to the effects of horizontal pleiotropy, these proteins may be useful markers of MM risk. Future work should explore the utility of these proteins in disease prediction or prevention using proteomic data from patients with MM or precursor conditions.

Reassessing the link between adiposity and head and neck cancer: a Mendelian randomization study

Background

Adiposity has been associated with an increased risk of head and neck cancer (HNC). Although body mass index (BMI) has been inversely associated with HNC risk among smokers, this is likely due to confounding. Previous Mendelian randomization (MR) studies could not fully discount causality between adiposity and HNC due to limited statistical power. Hence, we aimed to revisit this using the largest genome-wide association study (GWAS) of HNC available, which has more granular data on HNC subsites.

Methods

We assessed the genetically predicted effects of BMI (N=806,834), waist-to-hip ratio (WHR; N=697,734) and waist circumference (N=462,166) on the risk of HNC (N=12,264 cases and 19,259 controls) and its subsites (oral, laryngeal, hypopharyngeal and oropharyngeal cancers) using a two-sample MR framework. We used the inverse variance weighted (IVW) MR approach and multiple sensitivity analyses including the weighted median, weighted mode, MR-Egger, MR-PRESSO, and CAUSE approaches. We also used multivariable MR (MVMR) to explore the direct effects of the adiposity measures on HNC, while accounting for smoking behaviour, a well-known HNC risk factor.

Results

In univariable MR, higher genetically predicted BMI increased the risk of overall HNC (IVW OR=1.17 per 1 standard deviation [1-SD] higher BMI, 95% CI 1.02-1.34, p=0.03), with no heterogeneity across subsites (Q p=0.78). However, the effect was not consistent in sensitivity analyses. The IVW effect was attenuated when smoking was included in the MVMR model (OR accounting for comprehensive smoking index=0.96 per 1-SD higher BMI, 95% CI 0.80-1.15, p=0.64) and CAUSE indicated the IVW results could be biased by correlated pleiotropy. Furthermore, we did not find a link between genetically predicted WHR (IVW OR=1.05 per 1-SD higher WHR, 95% CI 0.89-1.24, p=0.53) or waist circumference and HNC risk (IVW OR=1.01 per 1-SD higher waist circumference, 95% CI 0.85-1.21, p=0.87).

Conclusions

Our findings suggest that adiposity does not play a role in HNC risk.