Population Structure of UK Biobank and Ancient Eurasians Reveals Adaptation at Genes Influencing Blood Pressure

Biobanking with genetics shapes precision medicine and global health

Precision medicine provides patients with access to personally tailored treatments based on individual-level data. However, developing personalized therapies requires analyses with substantial statistical power to map genetic and epidemiologic associations that ultimately create models informing clinical decisions. As one solution, biobanks have emerged as large-scale, longitudinal cohort studies with long-term storage of biological specimens and health information, including electronic health records and participant survey responses. By providing access to individual-level data for genotype-phenotype mapping efforts, pharmacogenomic studies, polygenic risk score assessments and rare variant analyses, biobanks support ongoing and future precision medicine research. Notably, due in part to the geographical enrichment of biobanks in Western Europe and North America, European ancestries have become disproportionately over-represented in precision medicine research. Herein, we provide a genetics-focused review of biobanks from around the world that are in pursuit of supporting precision medicine. We discuss the limitations of their designs, ongoing efforts to diversify genomics research and strategies to maximize the benefits of research leveraging biobanks for all.

A method to estimate the contribution of rare coding variants to complex trait heritability

It has been postulated that rare coding variants (RVs; MAF < 0.01) contribute to the "missing" heritability of complex traits. We developed a framework, the Rare variant heritability (RARity) estimator, to assess RV heritability (h2RV) without assuming a particular genetic architecture. We applied RARity to 31 complex traits in the UK Biobank (n = 167,348) and showed that gene-level RV aggregation suffers from 79% (95% CI: 68-93%) loss of h2RV. Using unaggregated variants, 27 traits had h2RV > 5%, with height having the highest h2RV at 21.9% (95% CI: 19.0-24.8%). The total heritability, including common and rare variants, recovered pedigree-based estimates for 11 traits. RARity can estimate gene-level h2RV, enabling the assessment of gene-level characteristics and revealing 11, previously unreported, gene-phenotype relationships. Finally, we demonstrated that in silico pathogenicity prediction (variant-level) and gene-level annotations do not generally enrich for RVs that over-contribute to complex trait variance, and thus, innovative methods are needed to predict RV functionality.

Highly parameterized polygenic scores tend to overfit to population stratification via random effects

Polygenic scores (PGSs), increasingly used in clinical settings, frequently include many genetic variants, with performance typically peaking at thousands of variants. Such highly parameterized PGSs often include variants that do not pass a genome-wide significance threshold. We propose a mathematical perspective that renders the effects of many of these non-significant variants random rather than causal, with the randomness capturing population structure. We devise methods to assess variant effect randomness and population stratification bias. Applying these methods to 141 traits from the UK Biobank, we find that, for many PGSs, the effects of non-significant variants are considerably random, with the extent of randomness associated with the degree of overfitting to population structure of the discovery cohort. Our findings explain why highly parameterized PGSs simultaneously have superior cohort-specific performance and limited generalizability, suggesting the critical need for variant randomness tests in PGS evaluation. Supporting code and a dashboard are available at https://github.com/songlab-cal/StratPGS.

A multi-cohort genome-wide association study in African ancestry individuals reveals risk loci for primary open-angle glaucoma

Primary open-angle glaucoma (POAG), the leading cause of irreversible blindness worldwide, disproportionately affects individuals of African ancestry. We conducted a genome-wide association study (GWAS) for POAG in 11,275 individuals of African ancestry (6,003 cases; 5,272 controls). We detected 46 risk loci associated with POAG at genome-wide significance. Replication and post-GWAS analyses, including functionally informed fine-mapping, multiple trait co-localization, and in silico validation, implicated two previously undescribed variants (rs1666698 mapping to DBF4P2; rs34957764 mapping to ROCK1P1) and one previously associated variant (rs11824032 mapping to ARHGEF12) as likely causal. For individuals of African ancestry, a polygenic risk score (PRS) for POAG from our mega-analysis (African ancestry individuals) outperformed a PRS from summary statistics of a much larger GWAS derived from European ancestry individuals. This study quantifies the genetic architecture similarities and differences between African and non-African ancestry populations for this blinding disease.

The selection landscape and genetic legacy of ancient Eurasians

The Holocene (beginning around 12,000 years ago) encompassed some of the most significant changes in human evolution, with far-reaching consequences for the dietary, physical and mental health of present-day populations. Using a dataset of more than 1,600 imputed ancient genomes1, we modelled the selection landscape during the transition from hunting and gathering, to farming and pastoralism across West Eurasia. We identify key selection signals related to metabolism, including that selection at the FADS cluster began earlier than previously reported and that selection near the LCT locus predates the emergence of the lactase persistence allele by thousands of years. We also find strong selection in the HLA region, possibly due to increased exposure to pathogens during the Bronze Age. Using ancient individuals to infer local ancestry tracts in over 400,000 samples from the UK Biobank, we identify widespread differences in the distribution of Mesolithic, Neolithic and Bronze Age ancestries across Eurasia. By calculating ancestry-specific polygenic risk scores, we show that height differences between Northern and Southern Europe are associated with differential Steppe ancestry, rather than selection, and that risk alleles for mood-related phenotypes are enriched for Neolithic farmer ancestry, whereas risk alleles for diabetes and Alzheimer's disease are enriched for Western hunter-gatherer ancestry. Our results indicate that ancient selection and migration were large contributors to the distribution of phenotypic diversity in present-day Europeans.

Testing for associations between HbA1c levels, polygenic risk and brain health in UK Biobank (N = 39 283)

AIM

To investigate whether continuous HbA1c levels and HbA1c-polygenic risk scores (HbA1c-PRS) are significantly associated with worse brain health independent of type 2 diabetes (T2D) diagnosis (vs. not), by examining brain structure and cognitive test score phenotypes.

METHODS

Using UK Biobank data (n = 39 283), we tested whether HbA1c levels and/or HbA1c-PRS were associated with cognitive test scores and brain imaging phenotypes. We adjusted for confounders of age, sex, Townsend deprivation score, level of education, genotyping chip, eight genetic principal components, smoking, alcohol intake frequency, cholesterol medication, body mass index, T2D and apolipoprotein (APOE) e4 dosage.

RESULTS

We found an association between higher HbA1c levels and poorer performance on symbol digit substitution scores (standardized beta [β] = -0.022, P = .001) in the fully adjusted model. We also found an association between higher HbA1c levels and worse brain MRI phenotypes of grey matter (GM; fully-adjusted β = -0.026, P < .001), whole brain volume (β = -0.072, P = .0113) and a general factor of frontal lobe GM (β = -0.022, P < .001) in partially and fully adjusted models. HbA1c-PRS were significantly associated with GM volume in the fully adjusted model (β = -0.010, P = .0113); however, when adjusted for HbA1c levels, the association was not significant.

CONCLUSIONS

Our findings suggest that measured HbA1c is associated with poorer cognitive health, and that HbA1c-PRS do not add significant information to this.

Human GLP1R variants affecting GLP1R cell surface expression are associated with impaired glucose control and increased adiposity

The glucagon-like peptide 1 receptor (GLP1R) is a major drug target with several agonists being prescribed in individuals with type 2 diabetes and obesity1,2. The impact of genetic variability of GLP1R on receptor function and its association with metabolic traits are unclear with conflicting reports. Here, we show an unexpected diversity of phenotypes ranging from defective cell surface expression to complete or pathway-specific gain of function (GoF) and loss of function (LoF), after performing a functional profiling of 60 GLP1R variants across four signalling pathways. The defective insulin secretion of GLP1R LoF variants is rescued by allosteric GLP1R ligands or high concentrations of exendin-4/semaglutide in INS-1 823/3 cells. Genetic association studies in 200,000 participants from the UK Biobank show that impaired GLP1R cell surface expression contributes to poor glucose control and increased adiposity with increased glycated haemoglobin A1c and body mass index. This study defines impaired GLP1R cell surface expression as a risk factor for traits associated with type 2 diabetes and obesity and provides potential treatment options for GLP1R LoF variant carriers.

MATS: a novel multi-ancestry transcriptome-wide association study to account for heterogeneity in the effects of cis-regulated gene expression on complex traits

The Transcriptome-Wide Association Study (TWAS) is a widely used approach which integrates gene expression and Genome Wide Association Study (GWAS) data to study the role of cis-regulated gene expression (GEx) in complex traits. However, the genetic architecture of GEx varies across populations, and recent findings point to possible ancestral heterogeneity in the effects of GEx on complex traits, which may be amplified in TWAS by modeling GEx as a function of cis-eQTLs. Here, we present a novel extension to TWAS to account for heterogeneity in the effects of cis-regulated GEx which are correlated with ancestry. Our proposed Multi-Ancestry TwaS (MATS) framework jointly analyzes samples from multiple populations and distinguishes between shared, ancestry-specific and/or subject-specific expression-trait associations. As such, MATS amplifies power to detect shared GEx associations over ancestry-stratified TWAS through increased sample sizes, and facilitates the detection of genes with subgroup-specific associations which may be masked by standard TWAS. Our simulations highlight the improved Type-I error conservation and power of MATS compared with competing approaches. Our real data applications to Alzheimer's disease (AD) case-control genotypes from the Alzheimer's Disease Sequencing Project (ADSP) and continuous phenotypes from the UK Biobank (UKBB) identify a number of unique gene-trait associations which were not discovered through standard and/or ancestry-stratified TWAS. Ultimately, these findings promote MATS as a powerful method for detecting and estimating significant gene expression effects on complex traits within multi-ancestry cohorts and corroborates the mounting evidence for inter-population heterogeneity in gene-trait associations.

Pathogen exposure misclassification can bias association signals in GWAS of infectious diseases when using population-based common control subjects

Genome-wide association studies (GWASs) have been performed to identify host genetic factors for a range of phenotypes, including for infectious diseases. The use of population-based common control subjects from biobanks and extensive consortia is a valuable resource to increase sample sizes in the identification of associated loci with minimal additional expense. Non-differential misclassification of the outcome has been reported when the control subjects are not well characterized, which often attenuates the true effect size. However, for infectious diseases the comparison of affected subjects to population-based common control subjects regardless of pathogen exposure can also result in selection bias. Through simulated comparisons of pathogen-exposed cases and population-based common control subjects, we demonstrate that not accounting for pathogen exposure can result in biased effect estimates and spurious genome-wide significant signals. Further, the observed association can be distorted depending upon strength of the association between a locus and pathogen exposure and the prevalence of pathogen exposure. We also used a real data example from the hepatitis C virus (HCV) genetic consortium comparing HCV spontaneous clearance to persistent infection with both well-characterized control subjects and population-based common control subjects from the UK Biobank. We find biased effect estimates for known HCV clearance-associated loci and potentially spurious HCV clearance associations. These findings suggest that the choice of control subjects is especially important for infectious diseases or outcomes that are conditional upon environmental exposures.

The missing link between genetic association and regulatory function

The genetic basis of most traits is highly polygenic and dominated by non-coding alleles. It is widely assumed that such alleles exert small regulatory effects on the expression of cis-linked genes. However, despite the availability of gene expression and epigenomic datasets, few variant-to-gene links have emerged. It is unclear whether these sparse results are due to limitations in available data and methods, or to deficiencies in the underlying assumed model. To better distinguish between these possibilities, we identified 220 gene-trait pairs in which protein-coding variants influence a complex trait or its Mendelian cognate. Despite the presence of expression quantitative trait loci near most GWAS associations, by applying a gene-based approach we found limited evidence that the baseline expression of trait-related genes explains GWAS associations, whether using colocalization methods (8% of genes implicated), transcription-wide association (2% of genes implicated), or a combination of regulatory annotations and distance (4% of genes implicated). These results contradict the hypothesis that most complex trait-associated variants coincide with homeostatic expression QTLs, suggesting that better models are needed. The field must confront this deficit and pursue this 'missing regulation.'

KLFDAPC: a supervised machine learning approach for spatial genetic structure analysis

Geographic patterns of human genetic variation provide important insights into human evolution and disease. A commonly used tool to detect and describe them is principal component analysis (PCA) or the supervised linear discriminant analysis of principal components (DAPC). However, genetic features produced from both approaches could fail to correctly characterize population structure for complex scenarios involving admixture. In this study, we introduce Kernel Local Fisher Discriminant Analysis of Principal Components (KLFDAPC), a supervised non-linear approach for inferring individual geographic genetic structure that could rectify the limitations of these approaches by preserving the multimodal space of samples. We tested the power of KLFDAPC to infer population structure and to predict individual geographic origin using neural networks. Simulation results showed that KLFDAPC has higher discriminatory power than PCA and DAPC. The application of our method to empirical European and East Asian genome-wide genetic datasets indicated that the first two reduced features of KLFDAPC correctly recapitulated the geography of individuals and significantly improved the accuracy of predicting individual geographic origin when compared to PCA and DAPC. Therefore, KLFDAPC can be useful for geographic ancestry inference, design of genome scans and correction for spatial stratification in GWAS that link genes to adaptation or disease susceptibility.

Efficient estimation of SNP heritability using Gaussian predictive process in large scale cohort studies

With the advent of high throughput genetic data, there have been attempts to estimate heritability from genome-wide SNP data on a cohort of distantly related individuals using linear mixed model (LMM). Fitting such an LMM in a large scale cohort study, however, is tremendously challenging due to its high dimensional linear algebraic operations. In this paper, we propose a new method named PredLMM approximating the aforementioned LMM motivated by the concepts of genetic coalescence and Gaussian predictive process. PredLMM has substantially better computational complexity than most of the existing LMM based methods and thus, provides a fast alternative for estimating heritability in large scale cohort studies. Theoretically, we show that under a model of genetic coalescence, the limiting form of our approximation is the celebrated predictive process approximation of large Gaussian process likelihoods that has well-established accuracy standards. We illustrate our approach with extensive simulation studies and use it to estimate the heritability of multiple quantitative traits from the UK Biobank cohort.

Association between polygenic risk for Alzheimer's disease, brain structure and cognitive abilities in UK Biobank

Previous studies testing associations between polygenic risk for late-onset Alzheimer's disease (LOAD-PGR) and brain magnetic resonance imaging (MRI) measures have been limited by small samples and inconsistent consideration of potential confounders. This study investigates whether higher LOAD-PGR is associated with differences in structural brain imaging and cognitive values in a relatively large sample of non-demented, generally healthy adults (UK Biobank). Summary statistics were used to create PGR scores for n = 32,790 participants using LDpred. Outcomes included 12 structural MRI volumes and 6 concurrent cognitive measures. Models were adjusted for age, sex, body mass index, genotyping chip, 8 genetic principal components, lifetime smoking, apolipoprotein (APOE) e4 genotype and socioeconomic deprivation. We tested for statistical interactions between APOE e4 allele dose and LOAD-PGR vs. all outcomes. In fully adjusted models, LOAD-PGR was associated with worse fluid intelligence (standardised beta [β] = -0.080 per LOAD-PGR standard deviation, p = 0.002), matrix completion (β = -0.102, p = 0.003), smaller left hippocampal total (β = -0.118, p = 0.002) and body (β = -0.069, p = 0.002) volumes, but not other hippocampal subdivisions. There were no significant APOE x LOAD-PGR score interactions for any outcomes in fully adjusted models. This is the largest study to date investigating LOAD-PGR and non-demented structural brain MRI and cognition phenotypes. LOAD-PGR was associated with smaller hippocampal volumes and aspects of cognitive ability in healthy adults and could supplement APOE status in risk stratification of cognitive impairment/LOAD.

Large-scale migration into Britain during the Middle to Late Bronze Age

Present-day people from England and Wales harbour more ancestry derived from Early European Farmers (EEF) than people of the Early Bronze Age¹. To understand this, we generated genome-wide data from 793 individuals, increasing data from the Middle to Late Bronze and Iron Age in Britain by 12-fold, and Western and Central Europe by 3.5-fold. Between 1000 and 875 BC, EEF ancestry increased in southern Britain (England and Wales) but not northern Britain (Scotland) due to incorporation of migrants who arrived at this time and over previous centuries, and who were genetically most similar to ancient individuals from France. These migrants contributed about half the ancestry of Iron Age people of England and Wales, thereby creating a plausible vector for the spread of early Celtic languages into Britain. These patterns are part of a broader trend of EEF ancestry becoming more similar across central and western Europe in the Middle to Late Bronze Age, coincident with archaeological evidence of intensified cultural exchange^2-6. There was comparatively less gene flow from continental Europe during the Iron Age, and Britain's independent genetic trajectory is also reflected in the rise of the allele conferring lactase persistence to ~50% by this time compared to ~7% in central Europe where it rose rapidly in frequency only a millennium later. This suggests that dairy products were used in qualitatively different ways in Britain and in central Europe over this period.

Incorporating functional priors improves polygenic prediction accuracy in UK Biobank and 23andMe data sets

Polygenic risk prediction is a widely investigated topic because of its promising clinical applications. Genetic variants in functional regions of the genome are enriched for complex trait heritability. Here, we introduce a method for polygenic prediction, LDpred-funct, that leverages trait-specific functional priors to increase prediction accuracy. We fit priors using the recently developed baseline-LD model, including coding, conserved, regulatory, and LD-related annotations. We analytically estimate posterior mean causal effect sizes and then use cross-validation to regularize these estimates, improving prediction accuracy for sparse architectures. We applied LDpred-funct to predict 21 highly heritable traits in the UK Biobank (avg N = 373 K as training data). LDpred-funct attained a +4.6% relative improvement in average prediction accuracy (avg prediction R2 = 0.144; highest R2 = 0.413 for height) compared to SBayesR (the best method that does not incorporate functional information). For height, meta-analyzing training data from UK Biobank and 23andMe cohorts (N = 1107 K) increased prediction R2 to 0.431. Our results show that incorporating functional priors improves polygenic prediction accuracy, consistent with the functional architecture of complex traits.

Genetically Increased Telomere Length and Aging-Related Traits in the U.K. Biobank

Telomere length (TL) shortens over time in most human cell types and is a potential biomarker of aging. However, the causal association of TL on physical and cognitive traits that decline with age has not been extensively examined in middle-aged adults. Using a Mendelian randomization (MR) approach, we utilized genetically increased TL (GI-TL) to estimate the impact of TL on aging-related traits among U.K. Biobank (UKB) participants (age 40-69 years). We manually curated 53 aging-related traits from the UKB and restricted to unrelated participants of British ancestry (n = 337,522). We estimated GI-TL as a linear combination of nine TL-associated single nucleotide polymorphisms (SNPs), each weighted by its previously-reported association with leukocyte TL. Regression models were used to assess the associations between GI-TL and each trait. We obtained MR estimates using the two-sample inverse variance weighted (IVW) approach. We identified six age-related traits associated with GI-TL (Bonferroni-corrected threshold p < .001): pulse pressure (PP) (p = 5.2 × 10-14), systolic blood pressure (SBP) (p = 2.9 × 10-15), diastolic blood pressure (DBP) (p = 5.5 × 10-6), hypertension (p = 5.5 × 10-11), forced expiratory volume (FEV1) (p = .0001), and forced vital capacity (FVC) (p = 3.8 × 10-6). Under MR assumptions, one standard deviation increase in TL (~1,200 base pairs) increased PP, SBP, and DBP by 1.5, 2.3, and 0.8 mmHg, respectively, while FEV1 and FVC increased by 34.7 and 52.2 mL, respectively. The observed associations appear unlikely to be due to selection bias based on analyses including inverse probability weights and analyses of simulated data. These findings suggest that longer TL increases pulmonary function and blood pressure traits among middle-aged UKB participants.

Deep sequencing of 1320 genes reveals the landscape of protein-truncating variants and their contribution to psoriasis in 19,973 Chinese individuals

Protein-truncating variants (PTVs) have important impacts on phenotype diversity and disease. However, their population genetics characteristics in more globally diverse populations are not well defined. Here, we describe patterns of PTVs in 1320 genes sequenced in 10,539 healthy controls and 9434 patients with psoriasis, all of Han Chinese ancestry. We identify 8720 PTVs, of which 77% are novel, and estimate 88% of all PTVs are deleterious and subject to purifying selection. Furthermore, we show that individuals with psoriasis have a significantly higher burden of PTVs compared to controls (P = 0.02). Finally, we identified 18 PTVs in 14 genes with unusually high levels of population differentiation, consistent with the action of local adaptation. Our study provides insights into patterns and consequences of PTVs.

Population genetic considerations for using biobanks as international resources in the pandemic era and beyond

The past years have seen the rise of genomic biobanks and mega-scale meta-analysis of genomic data, which promises to reveal the genetic underpinnings of health and disease. However, the over-representation of Europeans in genomic studies not only limits the global understanding of disease risk but also inhibits viable research into the genomic differences between carriers and patients. Whilst the community has agreed that more diverse samples are required, it is not enough to blindly increase diversity; the diversity must be quantified, compared and annotated to lead to insight. Genetic annotations from separate biobanks need to be comparable and computable and to operate without access to raw data due to privacy concerns. Comparability is key both for regular research and to allow international comparison in response to pandemics. Here, we evaluate the appropriateness of the most common genomic tools used to depict population structure in a standardized and comparable manner. The end goal is to reduce the effects of confounding and learn from genuine variation in genetic effects on phenotypes across populations, which will improve the value of biobanks (locally and internationally), increase the accuracy of association analyses and inform developmental efforts.

Evaluating individual genome similarity with a topic model

MOTIVATION

Evaluating genome similarity among individuals is an essential step in data analysis. Advanced sequencing technology detects more and rarer variants for massive individual genomes, thus enabling individual-level genome similarity evaluation. However, the current methodologies, such as the principal component analysis (PCA), lack the capability to fully leverage rare variants and are also difficult to interpret in terms of population genetics.

RESULTS

Here, we introduce a probabilistic topic model, latent Dirichlet allocation, to evaluate individual genome similarity. A total of 2535 individuals from the 1000 Genomes Project (KGP) were used to demonstrate our method. Various aspects of variant choice and model parameter selection were studied. We found that relatively rare (0.001<allele frequency < 0.175) and sparse (average interval > 20 000 bp) variants are more efficient for genome similarity evaluation. At least 100 000 such variants are necessary. In our results, the populations show significantly less mixed and more cohesive visualization than the PCA results. The global similarities among the KGP genomes are consistent with known geographical, historical and cultural factors.

AVAILABILITY AND IMPLEMENTATION

The source code and data access are available at: https://github.com/lrjuan/LDA_genome.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Human adaptation over the past 40,000 years

Over the past few years several methodological and data-driven advances have greatly improved our ability to robustly detect genomic signatures of selection in humans. New methods applied to large samples of present-day genomes provide increased power, while ancient DNA allows precise estimation of timing and tempo. However, despite these advances, we are still limited in our ability to translate these signatures into understanding about which traits were actually under selection, and why. Combining information from different populations and timescales may allow interpretation of selective sweeps. Other modes of selection have proved more difficult to detect. In particular, despite strong evidence of the polygenicity of most human traits, evidence for polygenic selection is weak, and its importance in recent human evolution remains unclear. Balancing selection and archaic introgression seem important for the maintenance of potentially adaptive immune diversity, but perhaps less so for other traits.

Genotyping Array Design and Data Quality Control in the Million Veteran Program

The Million Veteran Program (MVP), initiated by the Department of Veterans Affairs (VA), aims to collect biosamples with consent from at least one million veterans. Presently, blood samples have been collected from over 800,000 enrolled participants. The size and diversity of the MVP cohort, as well as the availability of extensive VA electronic health records, make it a promising resource for precision medicine. MVP is conducting array-based genotyping to provide a genome-wide scan of the entire cohort, in parallel with whole-genome sequencing, methylation, and other 'omics assays. Here, we present the design and performance of the MVP 1.0 custom Axiom array, which was designed and developed as a single assay to be used across the multi-ethnic MVP cohort. A unified genetic quality-control analysis was developed and conducted on an initial tranche of 485,856 individuals, leading to a high-quality dataset of 459,777 unique individuals. 668,418 genetic markers passed quality control and showed high-quality genotypes not only on common variants but also on rare variants. We confirmed that, with non-European individuals making up nearly 30%, MVP's substantial ancestral diversity surpasses that of other large biobanks. We also demonstrated the quality of the MVP dataset by replicating established genetic associations with height in European Americans and African Americans ancestries. This current dataset has been made available to approved MVP researchers for genome-wide association studies and other downstream analyses. Further data releases will be available for analysis as recruitment at the VA continues and the cohort expands both in size and diversity.

Polygenic adaptation on height is overestimated due to uncorrected stratification in genome-wide association studies

Genetic predictions of height differ among human populations and these differences have been interpreted as evidence of polygenic adaptation. These differences were first detected using SNPs genome-wide significantly associated with height, and shown to grow stronger when large numbers of sub-significant SNPs were included, leading to excitement about the prospect of analyzing large fractions of the genome to detect polygenic adaptation for multiple traits. Previous studies of height have been based on SNP effect size measurements in the GIANT Consortium meta-analysis. Here we repeat the analyses in the UK Biobank, a much more homogeneously designed study. We show that polygenic adaptation signals based on large numbers of SNPs below genome-wide significance are extremely sensitive to biases due to uncorrected population stratification. More generally, our results imply that typical constructions of polygenic scores are sensitive to population stratification and that population-level differences should be interpreted with caution.

Editorial note

This article has been through an editorial process in which the authors decide how to respond to the issues raised during peer review. The Reviewing Editor's assessment is that all the issues have been addressed (see decision letter).

Quantification of frequency-dependent genetic architectures in 25 UK Biobank traits reveals action of negative selection

Understanding the role of rare variants is important in elucidating the genetic basis of human disease. Negative selection can cause rare variants to have larger per-allele effect sizes than common variants. Here, we develop a method to estimate the minor allele frequency (MAF) dependence of SNP effect sizes. We use a model in which per-allele effect sizes have variance proportional to [p(1 - p)]α, where p is the MAF and negative values of α imply larger effect sizes for rare variants. We estimate α for 25 UK Biobank diseases and complex traits. All traits produce negative α estimates, with best-fit mean of -0.38 (s.e. 0.02) across traits. Despite larger rare variant effect sizes, rare variants (MAF < 1%) explain less than 10% of total SNP-heritability for most traits analyzed. Using evolutionary modeling and forward simulations, we validate the α model of MAF-dependent trait effects and assess plausible values of relevant evolutionary parameters.

Genetic association and transcriptome integration identify contributing genes and tissues at cystic fibrosis modifier loci

Cystic Fibrosis (CF) exhibits morbidity in several organs, including progressive lung disease in all patients and intestinal obstruction at birth (meconium ileus) in ~15%. Individuals with the same causal CFTR mutations show variable disease presentation which is partly attributed to modifier genes. With >6,500 participants from the International CF Gene Modifier Consortium, genome-wide association investigation identified a new modifier locus for meconium ileus encompassing ATP12A on chromosome 13 (min p = 3.83x10(-10)); replicated loci encompassing SLC6A14 on chromosome X and SLC26A9 on chromosome 1, (min p<2.2x10(-16), 2.81x10(-11), respectively); and replicated a suggestive locus on chromosome 7 near PRSS1 (min p = 2.55x10(-7)). PRSS1 is exclusively expressed in the exocrine pancreas and was previously associated with non-CF pancreatitis with functional characterization demonstrating impact on PRSS1 gene expression. We thus asked whether the other meconium ileus modifier loci impact gene expression and in which organ. We developed and applied a colocalization framework called the Simple Sum (SS) that integrates regulatory and genetic association information, and also contrasts colocalization evidence across tissues or genes. The associated modifier loci colocalized with expression quantitative trait loci (eQTLs) for ATP12A (p = 3.35x10(-8)), SLC6A14 (p = 1.12x10(-10)) and SLC26A9 (p = 4.48x10(-5)) in the pancreas, even though meconium ileus manifests in the intestine. The meconium ileus susceptibility locus on chromosome X appeared shifted in location from a previously identified locus for CF lung disease severity. Using the SS we integrated the lung disease association locus with eQTLs from nasal epithelia of 63 CF participants and demonstrated evidence of colocalization with airway-specific regulation of SLC6A14 (p = 2.3x10(-4)). Cystic Fibrosis is realizing the promise of personalized medicine, and identification of the contributing organ and understanding of tissue specificity for a gene modifier is essential for the next phase of personalizing therapeutic strategies.

Distinguishing genetic correlation from causation across 52 diseases and complex traits

Mendelian randomization, a method to infer causal relationships, is confounded by genetic correlations reflecting shared etiology. We developed a model in which a latent causal variable mediates the genetic correlation; trait 1 is partially genetically causal for trait 2 if it is strongly genetically correlated with the latent causal variable, quantified using the genetic causality proportion. We fit this model using mixed fourth moments [Formula: see text] and [Formula: see text] of marginal effect sizes for each trait; if trait 1 is causal for trait 2, then SNPs affecting trait 1 (large [Formula: see text]) will have correlated effects on trait 2 (large α1α2), but not vice versa. In simulations, our method avoided false positives due to genetic correlations, unlike Mendelian randomization. Across 52 traits (average n = 331,000), we identified 30 causal relationships with high genetic causality proportion estimates. Novel findings included a causal effect of low-density lipoprotein on bone mineral density, consistent with clinical trials of statins in osteoporosis.

High-throughput inference of pairwise coalescence times identifies signals of selection and enriched disease heritability

Interest in reconstructing demographic histories has motivated the development of methods to estimate locus-specific pairwise coalescence times from whole-genome sequence data. Here we introduce a powerful new method, ASMC, that can estimate coalescence times using only SNP array data, and is orders of magnitude faster than previous approaches. We applied ASMC to detect recent positive selection in 113,851 phased British samples from the UK Biobank, and detected 12 genome-wide significant signals, including 6 novel loci. We also applied ASMC to sequencing data from 498 Dutch individuals to detect background selection at deeper time scales. We detected strong heritability enrichment in regions of high background selection in an analysis of 20 independent diseases and complex traits using stratified LD score regression, conditioned on a broad set of functional annotations (including other background selection annotations). These results underscore the widespread effects of background selection on the genetic architecture of complex traits.

Deep whole-genome sequencing reveals recent selection signatures linked to evolution and disease risk of Japanese

Understanding natural selection is crucial to unveiling evolution of modern humans. Here, we report natural selection signatures in the Japanese population using 2234 high-depth whole-genome sequence (WGS) data (25.9×). Using rare singletons, we identify signals of very recent selection for the past 2000–3000 years in multiple loci (ADH cluster, MHC region, BRAP-ALDH2, SERHL2). In large-scale genome-wide association study (GWAS) dataset (n = 171,176), variants with selection signatures show enrichment in heterogeneity of derived allele frequency spectra among the geographic regions of Japan, highlighted by two major regional clusters (Hondo and Ryukyu). While the selection signatures do not show enrichment in archaic hominin-derived genome sequences, they overlap with the SNPs associated with the modern human traits. The strongest overlaps are observed for the alcohol or nutrition metabolism-related traits. Our study illustrates the value of high-depth WGS to understand evolution and their relationship with disease risk.

Recent natural selection left signals in human genomes. Here, Okada et al. generate high-depth whole-genome sequence (WGS) data (25.9×) from 2,234 Japanese people of the BioBank Japan Project (BBJ), and identify signals of recent natural selection which overlap variants associated with human traits.

Identifying the favored mutation in a positive selective sweep

Most approaches that capture signatures of selective sweeps in population genomics data do not identify the specific mutation favored by selection. We present iSAFE (for "integrated selection of allele favored by evolution"), a method that enables researchers to accurately pinpoint the favored mutation in a large region (∼5 Mbp) by using a statistic derived solely from population genetics signals. iSAFE does not require knowledge of demography, the phenotype under selection, or functional annotations of mutations.

Predicting age by mining electronic medical records with deep learning characterizes differences between chronological and physiological age

Determining the discrepancy between chronological and physiological age of patients is central to preventative and personalized care. Electronic medical records (EMR) provide rich information about the patient physiological state, but it is unclear whether such information can be predictive of chronological age. Here we present a deep learning model that uses vital signs and lab tests contained within the EMR of Mount Sinai Health System (MSHS) to predict chronological age. The model is trained on 377,686 EMR from patients of ages 18-85 years old. The discrepancy between the predicted and real chronological age is then used as a proxy to estimate physiological age. Overall, the model can predict the chronological age of patients with a standard deviation error of ∼7 years. The ages of the youngest and oldest patients were more accurately predicted, while patients of ages ranging between 40 and 60 years were the least accurately predicted. Patients with the largest discrepancy between their physiological and chronological age were further inspected. The patients predicted to be significantly older than their chronological age have higher systolic blood pressure, higher cholesterol, damaged liver, and anemia. In contrast, patients predicted to be younger than their chronological age have lower blood pressure and shorter stature among other indicators; both groups display lower weight than the population average. Using information from ∼10,000 patients from the entire cohort who have been also profiled with SNP arrays, genome-wide association study (GWAS) uncovers several novel genetic variants associated with aging. In particular, significant variants were mapped to genes known to be associated with inflammation, hypertension, lipid metabolism, height, and increased lifespan in mice. Several genes with missense mutations were identified as novel candidate aging genes. In conclusion, we demonstrate how EMR data can be used to assess overall health via a scale that is based on deviation from the patient's predicted chronological age.

Multiethnic polygenic risk scores improve risk prediction in diverse populations

Methods for genetic risk prediction have been widely investigated in recent years. However, most available training data involves European samples, and it is currently unclear how to accurately predict disease risk in other populations. Previous studies have used either training data from European samples in large sample size or training data from the target population in small sample size, but not both. Here, we introduce a multiethnic polygenic risk score that combines training data from European samples and training data from the target population. We applied this approach to predict type 2 diabetes (T2D) in a Latino cohort using both publicly available European summary statistics in large sample size (Neff  = 40k) and Latino training data in small sample size (Neff  = 8k). Here, we attained a >70% relative improvement in prediction accuracy (from R2  = 0.027 to 0.047) compared to methods that use only one source of training data, consistent with large relative improvements in simulations. We observed a systematically lower load of T2D risk alleles in Latino individuals with more European ancestry, which could be explained by polygenic selection in ancestral European and/or Native American populations. We predict T2D in a South Asian UK Biobank cohort using European (Neff  = 40k) and South Asian (Neff  = 16k) training data and attained a >70% relative improvement in prediction accuracy, and application to predict height in an African UK Biobank cohort using European (N = 113k) and African (N = 2k) training data attained a 30% relative improvement. Our work reduces the gap in polygenic risk prediction accuracy between European and non-European target populations.

Identifying genetic variants that affect viability in large cohorts

A number of open questions in human evolutionary genetics would become tractable if we were able to directly measure evolutionary fitness. As a step towards this goal, we developed a method to examine whether individual genetic variants, or sets of genetic variants, currently influence viability. The approach consists in testing whether the frequency of an allele varies across ages, accounting for variation in ancestry. We applied it to the Genetic Epidemiology Research on Adult Health and Aging (GERA) cohort and to the parents of participants in the UK Biobank. Across the genome, we found only a few common variants with large effects on age-specific mortality: tagging the APOE ε4 allele and near CHRNA3. These results suggest that when large, even late-onset effects are kept at low frequency by purifying selection. Testing viability effects of sets of genetic variants that jointly influence 1 of 42 traits, we detected a number of strong signals. In participants of the UK Biobank of British ancestry, we found that variants that delay puberty timing are associated with a longer parental life span (P~6.2 × 10⁻⁶ for fathers and P~2.0 × 10⁻³ for mothers), consistent with epidemiological studies. Similarly, variants associated with later age at first birth are associated with a longer maternal life span (P~1.4 × 10⁻³). Signals are also observed for variants influencing cholesterol levels, risk of coronary artery disease (CAD), body mass index, as well as risk of asthma. These signals exhibit consistent effects in the GERA cohort and among participants of the UK Biobank of non-British ancestry. We also found marked differences between males and females, most notably at the CHRNA3 locus, and variants associated with risk of CAD and cholesterol levels. Beyond our findings, the analysis serves as a proof of principle for how upcoming biomedical data sets can be used to learn about selection effects in contemporary humans.

Our global understanding of adaptation in humans is limited to indirect statistical inferences from patterns of genetic variation, which are sensitive to past selection pressures. We introduced a method that allowed us to directly observe ongoing selection in humans by identifying genetic variants that affect survival to a given age (i.e., viability selection). We applied our approach to the GERA cohort and parents of the UK Biobank participants. We found viability effects of variants near the APOE and CHRNA3 genes, which are associated with the risk of Alzheimer disease and smoking behavior, respectively. We also tested for the joint effect of sets of genetic variants that influence quantitative traits. We uncovered an association between longer life span and genetic variants that delay puberty timing and age at first birth. We also detected detrimental effects of higher genetically predicted cholesterol levels, body mass index, risk of coronary artery disease (CAD), and risk of asthma on survival. Some of the observed effects differ between males and females, most notably those at the CHRNA3 gene and variants associated with risk of CAD and cholesterol levels. Beyond this application, our analysis shows how large biomedical data sets can be used to study natural selection in humans.