Prediction of atrial fibrillation and stroke using machine learning models in UK Biobank.

Objective: Atrial fibrillation (AF) is the most common cardiac arrythmia, and it is associated with increased risk for ischemic stroke, which is underestimated, as AF can be asymptomatic. The aim of this study was to develop optimal ML models for prediction of AF in the population, and secondly for ischemic stroke in AF patients. Methods: To develop ML models for prediction of 1) AF in the general population and 2) ischemic stroke in patients with AF we constructed XGBoost, LightGBM, Random Forest, Deep Neural Network, Support Vector Machine and Lasso penalised logistic regression models using UK-Biobank's extensive real-world clinical data, questionnaires, as well as biochemical and genetic data, and their predictive performances were compared. Ranking and contribution of the different features was assessed by SHapley Additive exPlanations (SHAP) analysis. The clinical tool CHA2DS2-VASc for prediction of ischemic stroke among AF patients, was used for comparison to the best performing ML model. Findings: The best performing model for AF prediction was LightGBM, with an area-under-the-roc-curve (AUROC) of 0.729 (95% confidence intervals (CI): 0.719, 0.738). The best performing model for ischemic stroke prediction in AF patients was XGBoost with AUROC of 0.631 (95% CI: 0.604, 0.657). The improved AUROC in the XGBoost model compared to CHA2DS2-VASc was statistically significant based on DeLong's test (p-value = 2.20E-06). In addition, the SHAP analysis showed that several peripheral blood biomarkers (e.g. creatinine, glycated haemoglobin, monocytes) were associated with ischemic stroke, which are not considered by CHA2DS2-VASc. Implications: The best performing ML models presented have the potential for clinical use, but further validation in independent studies is required. Our results endorse the incorporation of some routinely measured blood biomarkers for ischemic stroke prediction in AF patients.

The role of thyroid function in borderline personality disorder and schizophrenia: a Mendelian Randomisation study.

BACKGROUND: Genome-wide association studies have reported a genetic overlap between borderline personality disorder (BPD) and schizophrenia (SCZ). Epidemiologically, the direction and causality of the association between thyroid function and risk of BPD and SCZ are unclear. We aim to test whether genetically predicted variations in TSH and FT4 levels or hypothyroidism are associated with the risk of BPD and SCZ. METHODS: We employed Mendelian Randomisation (MR) analyses using genetic instruments associated with TSH and FT4 levels as well as hypothyroidism to examine the effects of genetically predicted thyroid function on BPD and SCZ risk. Bidirectional MR analyses were employed to investigate a potential reverse causal association. RESULTS: Genetically predicted higher FT4 was not associated with the risk of BPD (OR: 1.18; P = 0.60, IVW) or the risk of SCZ (OR: 0.93; P = 0.19, IVW). Genetically predicted higher TSH was not associated with the risk of BPD (OR: 1.11; P = 0.51, IVW) or SCZ (OR: 0.98, P = 0.55, IVW). Genetically predicted hypothyroidism was not associated with BPD or SCZ. We found no evidence for a reverse causal effect between BPD or SCZ on thyroid function. CONCLUSIONS: We report evidence for a null association between genetically predicted FT4, TSH or hypothyroidism with BPD or SCZ risk. There was no evidence for reverse causality.

Multi-trait analysis characterizes the genetics of thyroid function and identifies causal associations with clinical implications.

To date only a fraction of the genetic footprint of thyroid function has been clarified. We report a genome-wide association study meta-analysis of thyroid function in up to 271,040 individuals of European ancestry, including reference range thyrotropin (TSH), free thyroxine (FT4), free and total triiodothyronine (T3), proxies for metabolism (T3/FT4 ratio) as well as dichotomized high and low TSH levels. We revealed 259 independent significant associations for TSH (61% novel), 85 for FT4 (67% novel), and 62 novel signals for the T3 related traits. The loci explained 14.1%, 6.0%, 9.5% and 1.1% of the total variation in TSH, FT4, total T3 and free T3 concentrations, respectively. Genetic correlations indicate that TSH associated loci reflect the thyroid function determined by free T3, whereas the FT4 associations represent the thyroid hormone metabolism. Polygenic risk score and Mendelian randomization analyses showed the effects of genetically determined variation in thyroid function on various clinical outcomes, including cardiovascular risk factors and diseases, autoimmune diseases, and cancer. In conclusion, our results improve the understanding of thyroid hormone physiology and highlight the pleiotropic effects of thyroid function on various diseases.

Applications of machine and deep learning to thyroid cytology and histopathology: a review.

This review synthesises past research into how machine and deep learning can improve the cyto- and histopathology processing pipelines for thyroid cancer diagnosis. The current gold-standard preoperative technique of fine-needle aspiration cytology has high interobserver variability, often returns indeterminate samples and cannot reliably identify some pathologies; histopathology analysis addresses these issues to an extent, but it requires surgical resection of the suspicious lesions so cannot influence preoperative decisions. Motivated by these issues, as well as by the chronic shortage of trained pathologists, much research has been conducted into how artificial intelligence could improve current pipelines and reduce the pressure on clinicians. Many past studies have indicated the significant potential of automated image analysis in classifying thyroid lesions, particularly for those of papillary thyroid carcinoma, but these have generally been retrospective, so questions remain about both the practical efficacy of these automated tools and the realities of integrating them into clinical workflows. Furthermore, the nature of thyroid lesion classification is significantly more nuanced in practice than many current studies have addressed, and this, along with the heterogeneous nature of processing pipelines in different laboratories, means that no solution has proven itself robust enough for clinical adoption. There are, therefore, multiple avenues for future research: examine the practical implementation of these algorithms as pathologist decision-support systems; improve interpretability, which is necessary for developing trust with clinicians and regulators; and investigate multiclassification on diverse multicentre datasets, aiming for methods that demonstrate high performance in a process- and equipment-agnostic manner.

Height, Autoimmune Thyroid Disease, and Thyroid Cancer: A Mendelian Randomization Study.

Background: Increased height has been associated with increased risk of hypothyroidism or thyroid cancer in epidemiological studies. However, the potential causal association between height and hypothyroidism or thyroid cancer has not been thoroughly explored. Autoimmune thyroid disease (AITD) mainly presents as hypothyroidism, thus we aim to evaluate the causal relationship between height as exposure and its association with AITD or thyroid cancer. Methods: Mendelian randomization (MR) analyses were performed by using genetic instruments associated with height, which were selected from the largest genome-wide association meta-analysis for height in up to 5.4 million individuals. Summary-level data for AITD and thyroid cancer (including 30,234 and 3001 cases, respectively) were collected from the large number of available genome-wide association studies. Bidirectional MR was performed to test for reverse causal association between AITD and adult height. Results: MR analyses showed that increased genetically predicted height was associated with a 4% increased risk of AITD ([CI 1.02 to 1.07], p-value = 1.99E-03) per 1-standard deviation (SD) increase in genetically predicted height. The bidirectional MR did not show any causal association between AITD and adult height. Additionally, increased genetically predicted height was associated with 15% increased risk of thyroid cancer ([CI 1.07 to 1.23], p-value = 2.32E-04) per 1-SD increase in height. Sensitivity analysis confirmed the main results. Conclusions: This MR study showed that 1-SD increase in genetically predicted height was associated with increased risk of AITD and thyroid cancer. In contrast, there was no evidence of a causal association of genetically predicted AITD with height. These results could further aid in investigation of height-related pathways as a means of gaining new mechanistic insights into AITD and thyroid cancer.

Identification and analysis of individuals who deviate from their genetically-predicted phenotype.

Findings from genome-wide association studies have facilitated the generation of genetic predictors for many common human phenotypes. Stratifying individuals misaligned to a genetic predictor based on common variants may be important for follow-up studies that aim to identify alternative causal factors. Using genome-wide imputed genetic data, we aimed to classify 158,951 unrelated individuals from the UK Biobank as either concordant or deviating from two well-measured phenotypes. We first applied our methods to standing height: our primary analysis classified 244 individuals (0.15%) as misaligned to their genetically predicted height. We show that these individuals are enriched for self-reporting being shorter or taller than average at age 10, diagnosed congenital malformations, and rare loss-of-function variants in genes previously catalogued as causal for growth disorders. Secondly, we apply our methods to LDL cholesterol (LDL-C). We classified 156 (0.12%) individuals as misaligned to their genetically predicted LDL-C and show that these individuals were enriched for both clinically actionable cardiovascular risk factors and rare genetic variants in genes previously shown to be involved in metabolic processes. Individuals whose LDL-C was higher than expected based on the genetic predictor were also at higher risk of developing coronary artery disease and type-two diabetes, even after adjustment for measured LDL-C, BMI and age, suggesting upward deviation from genetically predicted LDL-C is indicative of generally poor health. Our results remained broadly consistent when performing sensitivity analysis based on a variety of parametric and non-parametric methods to define individuals deviating from polygenic expectation. Our analyses demonstrate the potential importance of quantitatively identifying individuals for further follow-up based on deviation from genetic predictions.

Genetics of skeletal proportions in two different populations.

Human height can be divided into sitting height and leg length, reflecting growth of different parts of the skeleton whose relative proportions are captured by the ratio of sitting to total height (as sitting height ratio, SHR). Height is a highly heritable trait, and its genetic basis has been well-studied. However, the genetic determinants of skeletal proportion are much less well-characterized. Expanding substantially on past work, we performed a genome-wide association study (GWAS) of SHR in ∼450,000 individuals with European ancestry and ∼100,000 individuals with East Asian ancestry from the UK and China Kadoorie Biobanks. We identified 565 loci independently associated with SHR, including all genomic regions implicated in prior GWAS in these ancestries. While SHR loci largely overlap height-associated loci (P < 0.001), the fine-mapped SHR signals were often distinct from height. We additionally used fine-mapped signals to identify 36 credible sets with heterogeneous effects across ancestries. Lastly, we used SHR, sitting height, and leg length to identify genetic variation acting on specific body regions rather than on overall human height.

Genome-wide CRISPR screening of chondrocyte maturation newly implicates genes in skeletal growth and height-associated GWAS loci.

Alterations in the growth and maturation of chondrocytes can lead to variation in human height, including monogenic disorders of skeletal growth. We aimed to identify genes and pathways relevant to human growth by pairing human height genome-wide association studies (GWASs) with genome-wide knockout (KO) screens of growth-plate chondrocyte proliferation and maturation in vitro. We identified 145 genes that alter chondrocyte proliferation and maturation at early and/or late time points in culture, with 90% of genes validating in secondary screening. These genes are enriched in monogenic growth disorder genes and in KEGG pathways critical for skeletal growth and endochondral ossification. Further, common variants near these genes capture height heritability independent of genes computationally prioritized from GWASs. Our study emphasizes the value of functional studies in biologically relevant tissues as orthogonal datasets to refine likely causal genes from GWASs and implicates new genetic regulators of chondrocyte proliferation and maturation.

Robust Bioinformatics Approaches Result in the First Polygenic Risk Score for BMI in Greek Adults.

Quantifying the role of genetics via construction of polygenic risk scores (PRSs) is deemed a resourceful tool to enable and promote effective obesity prevention strategies. The present paper proposes a novel methodology for PRS extraction and presents the first PRS for body mass index (BMI) in a Greek population. A novel pipeline for PRS derivation was used to analyze genetic data from a unified database of three cohorts of Greek adults. The pipeline spans various steps of the process, from iterative dataset splitting to training and test partitions, calculation of summary statistics and PRS extraction, up to PRS aggregation and stabilization, achieving higher evaluation metrics. Using data from 2185 participants, implementation of the pipeline enabled consecutive repetitions in splitting training and testing samples and resulted in a 343-single nucleotide polymorphism PRS yielding an R2 = 0.3241 (beta = 1.011, p-value = 4 × 10-193) for BMI. PRS-included variants displayed a variety of associations with known traits (i.e., blood cell count, gut microbiome, lifestyle parameters). The proposed methodology led to creation of the first-ever PRS for BMI in Greek adults and aims at promoting a facilitating approach to reliable PRS development and integration in healthcare practice.

Identification and analysis of individuals who deviate from their genetically-predicted phenotype.

Findings from genome-wide association studies have facilitated the generation of genetic predictors for many common human phenotypes. Stratifying individuals misaligned to a genetic predictor based on common variants may be important for follow-up studies that aim to identify alternative causal factors. Using genome-wide imputed genetic data, we aimed to classify 158,951 unrelated individuals from the UK Biobank as either concordant or deviating from two well-measured phenotypes. We first applied our methods to standing height: our primary analysis classified 244 individuals (0.15%) as misaligned to their genetically predicted height. We show that these individuals are enriched for self-reporting being shorter or taller than average at age 10, diagnosed congenital malformations, and rare loss-of-function variants in genes previously catalogued as causal for growth disorders. Secondly, we apply our methods to LDL cholesterol. We classified 156 (0.12%) individuals as misaligned to their genetically predicted LDL cholesterol and show that these individuals were enriched for both clinically actionable cardiovascular risk factors and rare genetic variants in genes previously shown to be involved in metabolic processes. Individuals whose LDL-C was higher than expected based on the genetic predictor were also at higher risk of developing coronary artery disease and type-two diabetes, even after adjustment for measured LDL-C, BMI and age, suggesting upward deviation from genetically predicted LDL-C is indicative of generally poor health. Our results remained broadly consistent when performing sensitivity analysis based on a variety of parametric and non-parametric methods to define individuals deviating from polygenic expectation. Our analyses demonstrate the potential importance of quantitatively identifying individuals for further follow-up based on deviation from genetic predictions. Author Summary: Human genetics is becoming increasingly useful to help predict human traits across a population owing to findings from large-scale genetic association studies and advances in the power of genetic predictors. This provides an opportunity to potentially identify individuals that deviate from genetic predictions for a common phenotype under investigation. For example, an individual may be genetically predicted to be tall, but be shorter than expected. It is potentially important to identify individuals who deviate from genetic predictions as this can facilitate further follow-up to assess likely causes. Using 158,951 unrelated individuals from the UK Biobank, with height and LDL cholesterol, as exemplar traits, we demonstrate that approximately 0.15% & 0.12% of individuals deviate from their genetically predicted phenotypes respectively. We observed these individuals to be enriched for a range of rare clinical diagnoses, as well as rare genetic factors that may be causal. Our analyses also demonstrate several methods for detecting individuals who deviate from genetic predictions that can be applied to a range of continuous human phenotypes.

Implicating genes, pleiotropy, and sexual dimorphism at blood lipid loci through multi-ancestry meta-analysis.

BACKGROUND: Genetic variants within nearly 1000 loci are known to contribute to modulation of blood lipid levels. However, the biological pathways underlying these associations are frequently unknown, limiting understanding of these findings and hindering downstream translational efforts such as drug target discovery. RESULTS: To expand our understanding of the underlying biological pathways and mechanisms controlling blood lipid levels, we leverage a large multi-ancestry meta-analysis (N = 1,654,960) of blood lipids to prioritize putative causal genes for 2286 lipid associations using six gene prediction approaches. Using phenome-wide association (PheWAS) scans, we identify relationships of genetically predicted lipid levels to other diseases and conditions. We confirm known pleiotropic associations with cardiovascular phenotypes and determine novel associations, notably with cholelithiasis risk. We perform sex-stratified GWAS meta-analysis of lipid levels and show that 3-5% of autosomal lipid-associated loci demonstrate sex-biased effects. Finally, we report 21 novel lipid loci identified on the X chromosome. Many of the sex-biased autosomal and X chromosome lipid loci show pleiotropic associations with sex hormones, emphasizing the role of hormone regulation in lipid metabolism. CONCLUSIONS: Taken together, our findings provide insights into the biological mechanisms through which associated variants lead to altered lipid levels and potentially cardiovascular disease risk.

A saturated map of common genetic variants associated with human height.

Common single-nucleotide polymorphisms (SNPs) are predicted to collectively explain 40-50% of phenotypic variation in human height, but identifying the specific variants and associated regions requires huge sample sizes1. Here, using data from a genome-wide association study of 5.4 million individuals of diverse ancestries, we show that 12,111 independent SNPs that are significantly associated with height account for nearly all of the common SNP-based heritability. These SNPs are clustered within 7,209 non-overlapping genomic segments with a mean size of around 90 kb, covering about 21% of the genome. The density of independent associations varies across the genome and the regions of increased density are enriched for biologically relevant genes. In out-of-sample estimation and prediction, the 12,111 SNPs (or all SNPs in the HapMap 3 panel2) account for 40% (45%) of phenotypic variance in populations of European ancestry but only around 10-20% (14-24%) in populations of other ancestries. Effect sizes, associated regions and gene prioritization are similar across ancestries, indicating that reduced prediction accuracy is likely to be explained by linkage disequilibrium and differences in allele frequency within associated regions. Finally, we show that the relevant biological pathways are detectable with smaller sample sizes than are needed to implicate causal genes and variants. Overall, this study provides a comprehensive map of specific genomic regions that contain the vast majority of common height-associated variants. Although this map is saturated for populations of European ancestry, further research is needed to achieve equivalent saturation in other ancestries.

A multi-layer functional genomic analysis to understand noncoding genetic variation in lipids.

A major challenge of genome-wide association studies (GWASs) is to translate phenotypic associations into biological insights. Here, we integrate a large GWAS on blood lipids involving 1.6 million individuals from five ancestries with a wide array of functional genomic datasets to discover regulatory mechanisms underlying lipid associations. We first prioritize lipid-associated genes with expression quantitative trait locus (eQTL) colocalizations and then add chromatin interaction data to narrow the search for functional genes. Polygenic enrichment analysis across 697 annotations from a host of tissues and cell types confirms the central role of the liver in lipid levels and highlights the selective enrichment of adipose-specific chromatin marks in high-density lipoprotein cholesterol and triglycerides. Overlapping transcription factor (TF) binding sites with lipid-associated loci identifies TFs relevant in lipid biology. In addition, we present an integrative framework to prioritize causal variants at GWAS loci, producing a comprehensive list of candidate causal genes and variants with multiple layers of functional evidence. We highlight two of the prioritized genes, CREBRF and RRBP1, which show convergent evidence across functional datasets supporting their roles in lipid biology.

A multi-population phenome-wide association study of genetically-predicted height in the Million Veteran Program.

BACKGROUND: Height has been associated with many clinical traits but whether such associations are causal versus secondary to confounding remains unclear in many cases. To systematically examine this question, we performed a Mendelian Randomization-Phenome-wide association study (MR-PheWAS) using clinical and genetic data from a national healthcare system biobank. METHODS AND FINDINGS: Analyses were performed using data from the US Veterans Affairs (VA) Million Veteran Program in non-Hispanic White (EA, n = 222,300) and non-Hispanic Black (AA, n = 58,151) adults in the US. We estimated height genetic risk based on 3290 height-associated variants from a recent European-ancestry genome-wide meta-analysis. We compared associations of measured and genetically-predicted height with phenome-wide traits derived from the VA electronic health record, adjusting for age, sex, and genetic principal components. We found 345 clinical traits associated with measured height in EA and an additional 17 in AA. Of these, 127 were associated with genetically-predicted height at phenome-wide significance in EA and 2 in AA. These associations were largely independent from body mass index. We confirmed several previously described MR associations between height and cardiovascular disease traits such as hypertension, hyperlipidemia, coronary heart disease (CHD), and atrial fibrillation, and further uncovered MR associations with venous circulatory disorders and peripheral neuropathy in the presence and absence of diabetes. As a number of traits associated with genetically-predicted height frequently co-occur with CHD, we evaluated effect modification by CHD status of genetically-predicted height associations with risk factors for and complications of CHD. We found modification of effects of MR associations by CHD status for atrial fibrillation/flutter but not for hypertension, hyperlipidemia, or venous circulatory disorders. CONCLUSIONS: We conclude that height may be an unrecognized but biologically plausible risk factor for several common conditions in adults. However, more studies are needed to reliably exclude horizontal pleiotropy as a driving force behind at least some of the MR associations observed in this study.

DNA methylation signature of chronic low-grade inflammation and its role in cardio-respiratory diseases.

We performed a multi-ethnic Epigenome Wide Association study on 22,774 individuals to describe the DNA methylation signature of chronic low-grade inflammation as measured by C-Reactive protein (CRP). We find 1,511 independent differentially methylated loci associated with CRP. These CpG sites show correlation structures across chromosomes, and are primarily situated in euchromatin, depleted in CpG islands. These genomic loci are predominantly situated in transcription factor binding sites and genomic enhancer regions. Mendelian randomization analysis suggests altered CpG methylation is a consequence of increased blood CRP levels. Mediation analysis reveals obesity and smoking as important underlying driving factors for changed CpG methylation. Finally, we find that an activated CpG signature significantly increases the risk for cardiometabolic diseases and COPD.

Thyroid function, pernicious anemia and erythropoiesis: a two-sample Mendelian randomization study.

Autoimmune thyroid disease (AITD) and pernicious anemia (PA) often coexist, but the directionality is unknown. In a two-sample Mendelian randomization (MR) analysis, using summary statistics from large genome-wide association studies (GWASs) in Europeans (N = 49 269-755 406), we examined the genetic associations between thyroid function, PA and markers of erythropoiesis. We performed inverse variance weighted random-effects MR, several sensitivity MR analyses, and bidirectional MR and MR Steiger for directionality. AITD and PA were associated bidirectionally (P ≤ 8 × 10-6). Neither euthyroid thyroid stimulating hormone (TSH) nor free thyroxine (FT4) were causally associated with PA. One standard deviation (SD) increase in euthyroid FT4 regulated by genetic variants in deiodinases 1 and 2 genes (DIO1/DIO2), corresponding to low-normal free triiodothyronine (FT3) levels, was causally associated with a pernicious/macrocytic anemia pattern, i.e. decreased erythrocyte counts (rank-based inverse normal transformed ß = -0,064 [95% confidence interval: -0,085, -0,044], P = 8 × 10-10) and hemoglobin (-0.028 [-0.051, -0.005], P = 0.02) and increased mean corpuscular hemoglobin (0.058 [0.025, 0.091], P = 5 × 10-4) and mean corpuscular volume levels (0.075 [0.052, 0.098], P = 1 × 10-8). Meanwhile, subclinical hyperthyroidism mirrored that pattern. AITD was causally associated with increased erythrocyte distribution width (P = 0.007) and decreased reticulocyte counts (P ≤ 0.02), whereas high-normal FT4 regulated by DIO1/DIO2 variants was causally associated with decreased bilirubin (-0.039 (-0.064, -0.013), P = 0.003). In conclusion, the bidirectional association between AITD and PA suggests a shared heritability for these two autoimmune diseases. AITD was causally associated with impaired erythropoiesis and not autoimmune hemolysis. Additionally, in euthyroid individuals, local regulation of thyroid hormones by deiodinases likely plays a role in erythropoiesis.

Genetic variation influencing DNA methylation provides insights into molecular mechanisms regulating genomic function.

We determined the relationships between DNA sequence variation and DNA methylation using blood samples from 3,799 Europeans and 3,195 South Asians. We identify 11,165,559 SNP-CpG associations (methylation quantitative trait loci (meQTL), P < 10-14), including 467,915 meQTL that operate in trans. The meQTL are enriched for functionally relevant characteristics, including shared chromatin state, High-throuhgput chromosome conformation interaction, and association with gene expression, metabolic variation and clinical traits. We use molecular interaction and colocalization analyses to identify multiple nuclear regulatory pathways linking meQTL loci to phenotypic variation, including UBASH3B (body mass index), NFKBIE (rheumatoid arthritis), MGA (blood pressure) and COMMD7 (white cell counts). For rs6511961 , chromatin immunoprecipitation followed by sequencing (ChIP-seq) validates zinc finger protein (ZNF)333 as the likely trans acting effector protein. Finally, we used interaction analyses to identify population- and lineage-specific meQTL, including rs174548 in FADS1, with the strongest effect in CD8+ T cells, thus linking fatty acid metabolism with immune dysregulation and asthma. Our study advances understanding of the potential pathways linking genetic variation to human phenotype.

The power of genetic diversity in genome-wide association studies of lipids.

Increased blood lipid levels are heritable risk factors of cardiovascular disease with varied prevalence worldwide owing to different dietary patterns and medication use1. Despite advances in prevention and treatment, in particular through reducing low-density lipoprotein cholesterol levels2, heart disease remains the leading cause of death worldwide3. Genome-wideassociation studies (GWAS) of blood lipid levels have led to important biological and clinical insights, as well as new drug targets, for cardiovascular disease. However, most previous GWAS4-23 have been conducted in European ancestry populations and may have missed genetic variants that contribute to lipid-level variation in other ancestry groups. These include differences in allele frequencies, effect sizes and linkage-disequilibrium patterns24. Here we conduct a multi-ancestry, genome-wide genetic discovery meta-analysis of lipid levels in approximately 1.65 million individuals, including 350,000 of non-European ancestries. We quantify the gain in studying non-European ancestries and provide evidence to support the expansion of recruitment of additional ancestries, even with relatively small sample sizes. We find that increasing diversity rather than studying additional individuals of European ancestry results in substantial improvements in fine-mapping functional variants and portability of polygenic prediction (evaluated in approximately 295,000 individuals from 7 ancestry groupings). Modest gains in the number of discovered loci and ancestry-specific variants were also achieved. As GWAS expand emphasis beyond the identification of genes and fundamental biology towards the use of genetic variants for preventive and precision medicine25, we anticipate that increased diversity of participants will lead to more accurate and equitable26 application of polygenic scores in clinical practice.

Thyroid Function and the Risk of Alzheimer's Disease: A Mendelian Randomization Study.

Background: Observational studies suggest an association between thyroid function and risk of dementia, but the causality and direction of these effects are unclear. We aim to test whether genetically predicted variation within the normal range of thyroid function and hypothyroidism is causally associated with the risk of Alzheimer's disease (AD). Methods: Mendelian randomization (MR) analyses using genetic instruments are associated with normal range thyrotropin (TSH) and free thyroxine (fT4) levels. Secondary analyses included investigation of the role of hypothyroidism. Bidirectional MR was conducted to address the presence of a potential reverse causal association. Summary statistics were obtained from the ThyroidOmics Consortium involving up to 119,715 individuals and the latest AD genome-wide association study data including up to 71,880 cases. Results: MR analyses show an association between increased genetically predicted normal range TSH levels and a decreased risk of AD (p = 0.02). One standard deviation increased normal range TSH levels were associated with a decreased risk of AD in individuals younger than 50 years old (p = 0.04). There was no evidence for a causal association between fT4 (p = 0.54) and AD. We did not identify any effect of the genetically predicted full range TSH levels (p = 0.06) or hypothyroidism (p = 0.23) with AD. Bidirectional MR did not show any effect of genetic predisposition to AD on TSH or fT4 levels. Conclusions: This MR study shows that increased levels of genetically predicted TSH within the normal range and in younger individuals are associated with a decreased risk of AD. We observed a marginal association between genetically predicted full range TSH and AD risk. There was no evidence for an effect between genetically predicted fT4 or hypothyroidism on AD. Future studies should clarify the underlying pathophysiological mechanisms.