Methylome-wide studies of six metabolic traits.

Exploring the molecular correlates of metabolic health measures may identify the shared and unique biological processes and pathways that they track. Here, we performed epigenome-wide association studies (EWASs) of six metabolic traits: body mass index (BMI), body fat percentage, waist-hip ratio (WHR), and blood-based measures of glucose, high-density lipoprotein (HDL) cholesterol, and total cholesterol. We considered blood-based DNA methylation (DNAm) from >750,000 CpG sites in over 17,000 volunteers from the Generation Scotland (GS) cohort. Linear regression analyses identified between 304 and 11,815 significant CpGs per trait at P<3.6×10-8, with 37 significant CpG sites across all six traits. Further, we performed a Bayesian EWAS that jointly models all CpGs simultaneously and conditionally on each other, as opposed to the marginal linear regression analyses. This identified between 3 and 27 CpGs with a posterior inclusion probability ≥ 0.95 across the six traits. Next, we used elastic net penalised regression to train epigenetic scores (EpiScores) of each trait in GS, which were then tested in the Lothian Birth Cohort 1936 (LBC1936; European ancestry) and Health for Life in Singapore (HELIOS; Indian-, Malay- and Chinese-ancestries). A maximum of 27.1% of the variance in BMI was explained by the BMI EpiScore in the subset of Malay-ancestry Singaporeans. Four metabolic EpiScores were associated with general cognitive function in LBC1936 in models adjusted for vascular risk factors (Standardised ßrange: 0.08 - 0.12, PFDR < 0.05). EpiScores of metabolic health are applicable across ancestries and can reflect differences in brain health.

Genetic insights into the age-specific biological mechanisms governing human ovarian aging.

There is currently little evidence that the genetic basis of human phenotype varies significantly across the lifespan. However, time-to-event phenotypes are understudied and can be thought of as reflecting an underlying hazard, which is unlikely to be constant through life when values take a broad range. Here, we find that 74% of 245 genome-wide significant genetic associations with age at natural menopause (ANM) in the UK Biobank show a form of age-specific effect. Nineteen of these replicated discoveries are identified only by our modeling framework, which determines the time dependency of DNA-variant age-at-onset associations without a significant multiple-testing burden. Across the range of early to late menopause, we find evidence for significantly different underlying biological pathways, changes in the signs of genetic correlations of ANM to health indicators and outcomes, and differences in inferred causal relationships. We find that DNA damage response processes only act to shape ovarian reserve and depletion for women of early ANM. Genetically mediated delays in ANM were associated with increased relative risk of breast cancer and leiomyoma at all ages and with high cholesterol and heart failure for late-ANM women. These findings suggest that a better understanding of the age dependency of genetic risk factor relationships among health indicators and outcomes is achievable through appropriate statistical modeling of large-scale biobank data.

Refining epigenetic prediction of chronological and biological age.

BACKGROUND: Epigenetic clocks can track both chronological age (cAge) and biological age (bAge). The latter is typically defined by physiological biomarkers and risk of adverse health outcomes, including all-cause mortality. As cohort sample sizes increase, estimates of cAge and bAge become more precise. Here, we aim to develop accurate epigenetic predictors of cAge and bAge, whilst improving our understanding of their epigenomic architecture. METHODS: First, we perform large-scale (N = 18,413) epigenome-wide association studies (EWAS) of chronological age and all-cause mortality. Next, to create a cAge predictor, we use methylation data from 24,674 participants from the Generation Scotland study, the Lothian Birth Cohorts (LBC) of 1921 and 1936, and 8 other cohorts with publicly available data. In addition, we train a predictor of time to all-cause mortality as a proxy for bAge using the Generation Scotland cohort (1214 observed deaths). For this purpose, we use epigenetic surrogates (EpiScores) for 109 plasma proteins and the 8 component parts of GrimAge, one of the current best epigenetic predictors of survival. We test this bAge predictor in four external cohorts (LBC1921, LBC1936, the Framingham Heart Study and the Women's Health Initiative study). RESULTS: Through the inclusion of linear and non-linear age-CpG associations from the EWAS, feature pre-selection in advance of elastic net regression, and a leave-one-cohort-out (LOCO) cross-validation framework, we obtain cAge prediction with a median absolute error equal to 2.3 years. Our bAge predictor was found to slightly outperform GrimAge in terms of the strength of its association to survival (HRGrimAge = 1.47 [1.40, 1.54] with p = 1.08 × 10-52, and HRbAge = 1.52 [1.44, 1.59] with p = 2.20 × 10-60). Finally, we introduce MethylBrowsR, an online tool to visualise epigenome-wide CpG-age associations. CONCLUSIONS: The integration of multiple large datasets, EpiScores, non-linear DNAm effects, and new approaches to feature selection has facilitated improvements to the blood-based epigenetic prediction of biological and chronological age.

Liability-scale heritability estimation for biobank studies of low-prevalence disease.

Theory for liability-scale models of the underlying genetic basis of complex disease provides an important way to interpret, compare, and understand results generated from biological studies. In particular, through estimation of the liability-scale heritability (LSH), liability models facilitate an understanding and comparison of the relative importance of genetic and environmental risk factors that shape different clinically important disease outcomes. Increasingly, large-scale biobank studies that link genetic information to electronic health records, containing hundreds of disease diagnosis indicators that mostly occur infrequently within the sample, are becoming available. Here, we propose an extension of the existing liability-scale model theory suitable for estimating LSH in biobank studies of low-prevalence disease. In a simulation study, we find that our derived expression yields lower mean square error (MSE) and is less sensitive to prevalence misspecification as compared to previous transformations for diseases with ≤2% population prevalence and LSH of ≤0.45, especially if the biobank sample prevalence is less than that of the wider population. Applying our expression to 13 diagnostic outcomes of ≤3% prevalence in the UK Biobank study revealed important differences in LSH obtained from the different theoretical expressions that impact the conclusions made when comparing LSH across disease outcomes. This demonstrates the importance of careful consideration for estimation and prediction of low-prevalence disease outcomes and facilitates improved inference of the underlying genetic basis of ≤2% population prevalence diseases, especially where biobank sample ascertainment results in a healthier sample population.

Haematological changes from conception to childbirth: An indicator of major pregnancy complications.

BACKGROUND: About 800 women die every day worldwide from pregnancy-related complications, including excessive blood loss, infections and high-blood pressure (World Health Organization, 2019). To improve screening for high-risk pregnancies, we set out to identify patterns of maternal hematological changes associated with future pregnancy complications. METHODS: Using mixed effects models, we established changes in 14 complete blood count (CBC) parameters for 1710 healthy pregnancies and compared them to measurements from 98 pregnancy-induced hypertension, 106 gestational diabetes and 339 postpartum hemorrhage cases. RESULTS: Results show interindividual variations, but good individual repeatability in CBC values during physiological pregnancies, allowing the identification of specific alterations in women with obstetric complications. For example, in women with uncomplicated pregnancies, haemoglobin count decreases of 0.12 g/L (95% CI -0.16, -0.09) significantly per gestation week (p value <.001). Interestingly, this decrease is three times more pronounced in women who will develop pregnancy-induced hypertension, with an additional decrease of 0.39 g/L (95% CI -0.51, -0.26). We also confirm that obstetric complications and white CBC predict the likelihood of giving birth earlier during pregnancy. CONCLUSION: We provide a comprehensive description of the associations between haematological changes through pregnancy and three major obstetric complications to support strategies for prevention, early-diagnosis and maternal care.

Improving GWAS discovery and genomic prediction accuracy in biobank data.

Genetically informed, deep-phenotyped biobanks are an important research resource and it is imperative that the most powerful, versatile, and efficient analysis approaches are used. Here, we apply our recently developed Bayesian grouped mixture of regressions model (GMRM) in the UK and Estonian Biobanks and obtain the highest genomic prediction accuracy reported to date across 21 heritable traits. When compared to other approaches, GMRM accuracy was greater than annotation prediction models run in the LDAK or LDPred-funct software by 15% (SE 7%) and 14% (SE 2%), respectively, and was 18% (SE 3%) greater than a baseline BayesR model without single-nucleotide polymorphism (SNP) markers grouped into minor allele frequency-linkage disequilibrium (MAF-LD) annotation categories. For height, the prediction accuracy R2 was 47% in a UK Biobank holdout sample, which was 76% of the estimated [Formula: see text]. We then extend our GMRM prediction model to provide mixed-linear model association (MLMA) SNP marker estimates for genome-wide association (GWAS) discovery, which increased the independent loci detected to 16,162 in unrelated UK Biobank individuals, compared to 10,550 from BoltLMM and 10,095 from Regenie, a 62 and 65% increase, respectively. The average [Formula: see text] value of the leading markers increased by 15.24 (SE 0.41) for every 1% increase in prediction accuracy gained over a baseline BayesR model across the traits. Thus, we show that modeling genetic associations accounting for MAF and LD differences among SNP markers, and incorporating prior knowledge of genomic function, is important for both genomic prediction and discovery in large-scale individual-level studies.

Genome- and epigenome-wide studies of plasma protein biomarkers for Alzheimer's disease implicate TBCA and TREM2 in disease risk.

Introduction: The levels of many blood proteins are associated with Alzheimer's disease (AD) or its pathological hallmarks. Elucidating the molecular factors that control circulating levels of these proteins may help to identify proteins associated with disease risk mechanisms. Methods: Genome-wide and epigenome-wide studies (nindividuals ≤1064) were performed on plasma levels of 282 AD-associated proteins, identified by a structured literature review. Bayesian penalized regression estimated contributions of genetic and epigenetic variation toward inter-individual differences in plasma protein levels. Mendelian randomization (MR) and co-localization tested associations between proteins and disease-related phenotypes. Results: Sixty-four independent genetic and 26 epigenetic loci were associated with 45 proteins. Novel findings included an association between plasma triggering receptor expressed on myeloid cells 2 (TREM2) levels and a polymorphism and cytosine-phosphate-guanine (CpG) site within the MS4A4A locus. Higher plasma tubulin-specific chaperone A (TBCA) and TREM2 levels were significantly associated with lower AD risk. Discussion: Our data inform the regulation of biomarker levels and their relationships with AD.

Blood-based epigenome-wide analyses of cognitive abilities.

BACKGROUND: Blood-based markers of cognitive functioning might provide an accessible way to track neurodegeneration years prior to clinical manifestation of cognitive impairment and dementia. RESULTS: Using blood-based epigenome-wide analyses of general cognitive function, we show that individual differences in DNA methylation (DNAm) explain 35.0% of the variance in general cognitive function (g). A DNAm predictor explains ~4% of the variance, independently of a polygenic score, in two external cohorts. It also associates with circulating levels of neurology- and inflammation-related proteins, global brain imaging metrics, and regional cortical volumes. CONCLUSIONS: As sample sizes increase, the ability to assess cognitive function from DNAm data may be informative in settings where cognitive testing is unreliable or unavailable.

Probabilistic inference of the genetic architecture underlying functional enrichment of complex traits.

We develop a Bayesian model (BayesRR-RC) that provides robust SNP-heritability estimation, an alternative to marker discovery, and accurate genomic prediction, taking 22 seconds per iteration to estimate 8.4 million SNP-effects and 78 SNP-heritability parameters in the UK Biobank. We find that only ≤10% of the genetic variation captured for height, body mass index, cardiovascular disease, and type 2 diabetes is attributable to proximal regulatory regions within 10kb upstream of genes, while 12-25% is attributed to coding regions, 32-44% to introns, and 22-28% to distal 10-500kb upstream regions. Up to 24% of all cis and coding regions of each chromosome are associated with each trait, with over 3,100 independent exonic and intronic regions and over 5,400 independent regulatory regions having ≥95% probability of contributing ≥0.001% to the genetic variance of these four traits. Our open-source software (GMRM) provides a scalable alternative to current approaches for biobank data.

Postpartum hemorrhage risk is driven by changes in blood composition through pregnancy.

The extent to which women differ in the course of blood cell counts throughout pregnancy, and the importance of these changes to pregnancy outcomes has not been well defined. Here, we develop a series of statistical analyses of repeated measures data to reveal the degree to which women differ in the course of pregnancy, predict the changes that occur, and determine the importance of these changes for post-partum hemorrhage (PPH) which is one of the leading causes of maternal mortality. We present a prospective cohort of 4082 births recorded at the University Hospital, Lausanne, Switzerland between 2009 and 2014 where full labour records could be obtained, along with complete blood count data taken at hospital admission. We find significant differences, at a [Formula: see text] level, among women in how blood count values change through pregnancy for mean corpuscular hemoglobin, mean corpuscular volume, mean platelet volume, platelet count and red cell distribution width. We find evidence that almost all complete blood count values show trimester-specific associations with PPH. For example, high platelet count (OR 1.20, 95% CI 1.01-1.53), high mean platelet volume (OR 1.58, 95% CI 1.04-2.08), and high erythrocyte levels (OR 1.36, 95% CI 1.01-1.57) in trimester 1 increased PPH, but high values in trimester 3 decreased PPH risk (OR 0.85, 0.79, 0.67 respectively). We show that differences among women in the course of blood cell counts throughout pregnancy have an important role in shaping pregnancy outcome and tracking blood count value changes through pregnancy improves identification of women at increased risk of postpartum hemorrhage. This study provides greater understanding of the complex changes in blood count values that occur through pregnancy and provides indicators to guide the stratification of patients into risk groups.

Genomic architecture and prediction of censored time-to-event phenotypes with a Bayesian genome-wide analysis.

While recent advancements in computation and modelling have improved the analysis of complex traits, our understanding of the genetic basis of the time at symptom onset remains limited. Here, we develop a Bayesian approach (BayesW) that provides probabilistic inference of the genetic architecture of age-at-onset phenotypes in a sampling scheme that facilitates biobank-scale time-to-event analyses. We show in extensive simulation work the benefits BayesW provides in terms of number of discoveries, model performance and genomic prediction. In the UK Biobank, we find many thousands of common genomic regions underlying the age-at-onset of high blood pressure (HBP), cardiac disease (CAD), and type-2 diabetes (T2D), and for the genetic basis of onset reflecting the underlying genetic liability to disease. Age-at-menopause and age-at-menarche are also highly polygenic, but with higher variance contributed by low frequency variants. Genomic prediction into the Estonian Biobank data shows that BayesW gives higher prediction accuracy than other approaches.

Author Correction: Bayesian reassessment of the epigenetic architecture of complex traits.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Multi-method genome- and epigenome-wide studies of inflammatory protein levels in healthy older adults.

BACKGROUND: The molecular factors which control circulating levels of inflammatory proteins are not well understood. Furthermore, association studies between molecular probes and human traits are often performed by linear model-based methods which may fail to account for complex structure and interrelationships within molecular datasets. METHODS: In this study, we perform genome- and epigenome-wide association studies (GWAS/EWAS) on the levels of 70 plasma-derived inflammatory protein biomarkers in healthy older adults (Lothian Birth Cohort 1936; n = 876; Olink® inflammation panel). We employ a Bayesian framework (BayesR+) which can account for issues pertaining to data structure and unknown confounding variables (with sensitivity analyses using ordinary least squares- (OLS) and mixed model-based approaches). RESULTS: We identified 13 SNPs associated with 13 proteins (n = 1 SNP each) concordant across OLS and Bayesian methods. We identified 3 CpG sites spread across 3 proteins (n = 1 CpG each) that were concordant across OLS, mixed-model and Bayesian analyses. Tagged genetic variants accounted for up to 45% of variance in protein levels (for MCP2, 36% of variance alone attributable to 1 polymorphism). Methylation data accounted for up to 46% of variation in protein levels (for CXCL10). Up to 66% of variation in protein levels (for VEGFA) was explained using genetic and epigenetic data combined. We demonstrated putative causal relationships between CD6 and IL18R1 with inflammatory bowel disease and between IL12B and Crohn's disease. CONCLUSIONS: Our data may aid understanding of the molecular regulation of the circulating inflammatory proteome as well as causal relationships between inflammatory mediators and disease.

Bayesian reassessment of the epigenetic architecture of complex traits.

Linking epigenetic marks to clinical outcomes improves insight into molecular processes, disease prediction, and therapeutic target identification. Here, a statistical approach is presented to infer the epigenetic architecture of complex disease, determine the variation captured by epigenetic effects, and estimate phenotype-epigenetic probe associations jointly. Implicitly adjusting for probe correlations, data structure (cell-count or relatedness), and single-nucleotide polymorphism (SNP) marker effects, improves association estimates and in 9,448 individuals, 75.7% (95% CI 71.70-79.3) of body mass index (BMI) variation and 45.6% (95% CI 37.3-51.9) of cigarette consumption variation was captured by whole blood methylation array data. Pathway-linked probes of blood cholesterol, lipid transport and sterol metabolism for BMI, and xenobiotic stimuli response for smoking, showed >1.5 times larger associations with >95% posterior inclusion probability. Prediction accuracy improved by 28.7% for BMI and 10.2% for smoking over a LASSO model, with age-, and tissue-specificity, implying associations are a phenotypic consequence rather than causal.

Quantification of the overall contribution of gene-environment interaction for obesity-related traits.

The growing sample size of genome-wide association studies has facilitated the discovery of gene-environment interactions (GxE). Here we propose a maximum likelihood method to estimate the contribution of GxE to continuous traits taking into account all interacting environmental variables, without the need to measure any. Extensive simulations demonstrate that our method provides unbiased interaction estimates and excellent coverage. We also offer strategies to distinguish specific GxE from general scale effects. Applying our method to 32 traits in the UK Biobank reveals that while the genetic risk score (GRS) of 376 variants explains 5.2% of body mass index (BMI) variance, GRSxE explains an additional 1.9%. Nevertheless, this interaction holds for any variable with identical correlation to BMI as the GRS, hence may not be GRS-specific. Still, we observe that the global contribution of specific GRSxE to complex traits is substantial for nine obesity-related measures (including leg impedance and trunk fat-free mass).

Significant out-of-sample classification from methylation profile scoring for amyotrophic lateral sclerosis.

We conducted DNA methylation association analyses using Illumina 450K data from whole blood for an Australian amyotrophic lateral sclerosis (ALS) case-control cohort (782 cases and 613 controls). Analyses used mixed linear models as implemented in the OSCA software. We found a significantly higher proportion of neutrophils in cases compared to controls which replicated in an independent cohort from the Netherlands (1159 cases and 637 controls). The OSCA MOMENT linear mixed model has been shown in simulations to best account for confounders. When combined in a methylation profile score, the 25 most-associated probes identified by MOMENT significantly classified case-control status in the Netherlands sample (area under the curve, AUC = 0.65, CI95% = [0.62-0.68], p = 8.3 × 10-22). The maximum AUC achieved was 0.69 (CI95% = [0.66-0.71], p = 4.3 × 10-34) when cell-type proportion was included in the predictor.

Accurate, scalable and integrative haplotype estimation.

The number of human genomes being genotyped or sequenced increases exponentially and efficient haplotype estimation methods able to handle this amount of data are now required. Here we present a method, SHAPEIT4, which substantially improves upon other methods to process large genotype and high coverage sequencing datasets. It notably exhibits sub-linear running times with sample size, provides highly accurate haplotypes and allows integrating external phasing information such as large reference panels of haplotypes, collections of pre-phased variants and long sequencing reads. We provide SHAPEIT4 in an open source format and demonstrate its performance in terms of accuracy and running times on two gold standard datasets: the UK Biobank data and the Genome In A Bottle.

Mitochondrial haplotypes affect metabolic phenotypes in the Drosophila Genetic Reference Panel.

The nature and extent of mitochondrial DNA variation in a population and how it affects traits is poorly understood. Here we resequence the mitochondrial genomes of 169 Drosophila Genetic Reference Panel lines, identifying 231 variants that stratify along 12 mitochondrial haplotypes. We identify 1,845 cases of mitonuclear allelic imbalances, thus implying that mitochondrial haplotypes are reflected in the nuclear genome. However, no major fitness effects are associated with mitonuclear imbalance, suggesting that such imbalances reflect population structure at the mitochondrial level rather than genomic incompatibilities. Although mitochondrial haplotypes have no direct impact on mitochondrial respiration, some haplotypes are associated with stress- and metabolism-related phenotypes, including food intake in males. Finally, through reciprocal swapping of mitochondrial genomes, we demonstrate that a mitochondrial haplotype associated with high food intake can rescue a low food intake phenotype. Together, our findings provide new insight into population structure at the mitochondrial level and point to the importance of incorporating mitochondrial haplotypes in genotype-phenotype relationship studies.

Dissection of genetic variation and evidence for pleiotropy in male pattern baldness.

Male pattern baldness (MPB) is a sex-limited, age-related, complex trait. We study MPB genetics in 205,327 European males from the UK Biobank. Here we show that MPB is strongly heritable and polygenic, with pedigree-heritability of 0.62 (SE = 0.03) estimated from close relatives, and SNP-heritability of 0.39 (SE = 0.01) from conventionally-unrelated males. We detect 624 near-independent genome-wide loci, contributing SNP-heritability of 0.25 (SE = 0.01), of which 26 X-chromosome loci explain 11.6%. Autosomal genetic variance is enriched for common variants and regions of lower linkage disequilibrium. We identify plausible genetic correlations between MPB and multiple sex-limited markers of earlier puberty, increased bone mineral density (rg = 0.15) and pancreatic ß-cell function (rg = 0.12). Correlations with reproductive traits imply an effect on fitness, consistent with an estimated linear selection gradient of -0.018 per MPB standard deviation. Overall, we provide genetic insights into MPB: a phenotype of interest in its own right, with value as a model sex-limited, complex trait.