Exploring machine learning strategies for predicting cardiovascular disease risk factors from multi-omic data.

BACKGROUND: Machine learning (ML) classifiers are increasingly used for predicting cardiovascular disease (CVD) and related risk factors using omics data, although these outcomes often exhibit categorical nature and class imbalances. However, little is known about which ML classifier, omics data, or upstream dimension reduction strategy has the strongest influence on prediction quality in such settings. Our study aimed to illustrate and compare different machine learning strategies to predict CVD risk factors under different scenarios. METHODS: We compared the use of six ML classifiers in predicting CVD risk factors using blood-derived metabolomics, epigenetics and transcriptomics data. Upstream omic dimension reduction was performed using either unsupervised or semi-supervised autoencoders, whose downstream ML classifier performance we compared. CVD risk factors included systolic and diastolic blood pressure measurements and ultrasound-based biomarkers of left ventricular diastolic dysfunction (LVDD; E/e' ratio, E/A ratio, LAVI) collected from 1,249 Finnish participants, of which 80% were used for model fitting. We predicted individuals with low, high or average levels of CVD risk factors, the latter class being the most common. We constructed multi-omic predictions using a meta-learner that weighted single-omic predictions. Model performance comparisons were based on the F1 score. Finally, we investigated whether learned omic representations from pre-trained semi-supervised autoencoders could improve outcome prediction in an external cohort using transfer learning. RESULTS: Depending on the ML classifier or omic used, the quality of single-omic predictions varied. Multi-omics predictions outperformed single-omics predictions in most cases, particularly in the prediction of individuals with high or low CVD risk factor levels. Semi-supervised autoencoders improved downstream predictions compared to the use of unsupervised autoencoders. In addition, median gains in Area Under the Curve by transfer learning compared to modelling from scratch ranged from 0.09 to 0.14 and 0.07 to 0.11 units for transcriptomic and metabolomic data, respectively. CONCLUSIONS: By illustrating the use of different machine learning strategies in different scenarios, our study provides a platform for researchers to evaluate how the choice of omics, ML classifiers, and dimension reduction can influence the quality of CVD risk factor predictions.

Comprehensive Inherited Risk Estimation for Risk-Based Breast Cancer Screening in Women.

PURPOSE: Family history (FH) and pathogenic variants (PVs) are used for guiding risk surveillance in selected high-risk women but little is known about their impact for breast cancer screening on population level. In addition, polygenic risk scores (PRSs) have been shown to efficiently stratify breast cancer risk through combining information about common genetic factors into one measure. METHODS: In longitudinal real-life data, we evaluate PRS, FH, and PVs for stratified screening. Using FinnGen (N = 117,252), linked to the Mass Screening Registry for breast cancer (1992-2019; nationwide organized biennial screening for age 50-69 years), we assessed the screening performance of a breast cancer PRS and compared its performance with FH of breast cancer and PVs in moderate- (CHEK2)- to high-risk (PALB2) susceptibility genes. RESULTS: Effect sizes for FH, PVs, and high PRS (>90th percentile) were comparable in screening-aged women, with similar implications for shifting age at screening onset. A high PRS identified women more likely to be diagnosed with breast cancer after a positive screening finding (positive predictive value [PPV], 39.5% [95% CI, 37.6 to 41.5]). Combinations of risk factors increased the PPVs up to 45% to 50%. A high PRS conferred an elevated risk of interval breast cancer (hazard ratio [HR], 2.78 [95% CI, 2.00 to 3.86] at age 50 years; HR, 2.48 [95% CI, 1.67 to 3.70] at age 60 years), and women with a low PRS (<10th percentile) had a low risk for both interval- and screen-detected breast cancers. CONCLUSION: Using real-life screening data, this study demonstrates the effectiveness of a breast cancer PRS for risk stratification, alone and combined with FH and PVs. Further research is required to evaluate their impact in a prospective risk-stratified screening program, including cost-effectiveness.

Disentangling the link between maternal influences on birth weight and disease risk in 36,211 genotyped mother-child pairs.

Epidemiological studies have robustly linked lower birth weight to later-life disease risks. These observations may reflect the adverse impact of intrauterine growth restriction on a child's health. However, causal evidence supporting such a mechanism in humans is largely lacking. Using Mendelian Randomization and 36,211 genotyped mother-child pairs from the FinnGen study, we assessed the relationship between intrauterine growth and five common health outcomes (coronary heart disease (CHD), hypertension, statin use, type 2 diabetes and cancer). We proxied intrauterine growth with polygenic scores for maternal effects on birth weight and took into account the transmission of genetic variants between a mother and a child in the analyses. We find limited evidence for contribution of normal variation in maternally influenced intrauterine growth on later-life disease. Instead, we find support for genetic pleiotropy in the fetal genome linking birth weight to CHD and hypertension. Our study illustrates the opportunities that data from genotyped parent-child pairs from a population-based biobank provides for addressing causality of maternal influences.

Distinct and shared genetic architectures of gestational diabetes mellitus and type 2 diabetes.

Gestational diabetes mellitus (GDM) is a common metabolic disorder affecting more than 16 million pregnancies annually worldwide1,2. GDM is related to an increased lifetime risk of type 2 diabetes (T2D)1-3, with over a third of women developing T2D within 15 years of their GDM diagnosis. The diseases are hypothesized to share a genetic predisposition1-7, but few studies have sought to uncover the genetic underpinnings of GDM. Most studies have evaluated the impact of T2D loci only8-10, and the three prior genome-wide association studies of GDM11-13 have identified only five loci, limiting the power to assess to what extent variants or biological pathways are specific to GDM. We conducted the largest genome-wide association study of GDM to date in 12,332 cases and 131,109 parous female controls in the FinnGen study and identified 13 GDM-associated loci, including nine new loci. Genetic features distinct from T2D were identified both at the locus and genomic scale. Our results suggest that the genetics of GDM risk falls into the following two distinct categories: one part conventional T2D polygenic risk and one part predominantly influencing mechanisms disrupted in pregnancy. Loci with GDM-predominant effects map to genes related to islet cells, central glucose homeostasis, steroidogenesis and placental expression.

Longitudinal multi-omics study reveals common etiology underlying association between plasma proteome and BMI trajectories in adolescent and young adult twins.

BACKGROUND: The influence of genetics and environment on the association of the plasma proteome with body mass index (BMI) and changes in BMI remains underexplored, and the links to other omics in these associations remain to be investigated. We characterized protein-BMI trajectory associations in adolescents and adults and how these connect to other omics layers. METHODS: Our study included two cohorts of longitudinally followed twins: FinnTwin12 (N = 651) and the Netherlands Twin Register (NTR) (N = 665). Follow-up comprised 4 BMI measurements over approximately 6 (NTR: 23-27 years old) to 10 years (FinnTwin12: 12-22 years old), with omics data collected at the last BMI measurement. BMI changes were calculated in latent growth curve models. Mixed-effects models were used to quantify the associations between the abundance of 439 plasma proteins with BMI at blood sampling and changes in BMI. In FinnTwin12, the sources of genetic and environmental variation underlying the protein abundances were quantified by twin models, as were the associations of proteins with BMI and BMI changes. In NTR, we investigated the association of gene expression of genes encoding proteins identified in FinnTwin12 with BMI and changes in BMI. We linked identified proteins and their coding genes to plasma metabolites and polygenic risk scores (PRS) applying mixed-effects models and correlation networks. RESULTS: We identified 66 and 14 proteins associated with BMI at blood sampling and changes in BMI, respectively. The average heritability of these proteins was 35%. Of the 66 BMI-protein associations, 43 and 12 showed genetic and environmental correlations, respectively, including 8 proteins showing both. Similarly, we observed 7 and 3 genetic and environmental correlations between changes in BMI and protein abundance, respectively. S100A8 gene expression was associated with BMI at blood sampling, and the PRG4 and CFI genes were associated with BMI changes. Proteins showed strong connections with metabolites and PRSs, but we observed no multi-omics connections among gene expression and other omics layers. CONCLUSIONS: Associations between the proteome and BMI trajectories are characterized by shared genetic, environmental, and metabolic etiologies. We observed few gene-protein pairs associated with BMI or changes in BMI at the proteome and transcriptome levels.

Genome-wide association analysis of plasma lipidome identifies 495 genetic associations.

The human plasma lipidome captures risk for cardiometabolic diseases. To discover new lipid-associated variants and understand the link between lipid species and cardiometabolic disorders, we perform univariate and multivariate genome-wide analyses of 179 lipid species in 7174 Finnish individuals. We fine-map the associated loci, prioritize genes, and examine their disease links in 377,277 FinnGen participants. We identify 495 genome-trait associations in 56 genetic loci including 8 novel loci, with a considerable boost provided by the multivariate analysis. For 26 loci, fine-mapping identifies variants with a high causal probability, including 14 coding variants indicating likely causal genes. A phenome-wide analysis across 953 disease endpoints reveals disease associations for 40 lipid loci. For 11 coronary artery disease risk variants, we detect strong associations with lipid species. Our study demonstrates the power of multivariate genetic analysis in correlated lipidomics data and reveals genetic links between diseases and lipid species beyond the standard lipids.

Longitudinal multi-omics study reveals common etiology underlying association between plasma proteome and BMI trajectories in adolescent and young adult twins.

Background: The influence of genetics and environment on the association of the plasma proteome with body mass index (BMI) and changes in BMI remain underexplored, and the links to other omics in these associations remain to be investigated. We characterized protein-BMI trajectory associations in adolescents and adults and how these connect to other omics layers. Methods: Our study included two cohorts of longitudinally followed twins: FinnTwin12 (N=651) and the Netherlands Twin Register (NTR) (N=665). Follow-up comprised four BMI measurements over approximately 6 (NTR: 23-27 years old) to 10 years (FinnTwin12: 12-22 years old), with omics data collected at the last BMI measurement. BMI changes were calculated using latent growth curve models. Mixed-effects models were used to quantify the associations between the abundance of 439 plasma proteins with BMI at blood sampling and changes in BMI. The sources of genetic and environmental variation underlying the protein abundances were quantified using twin models, as were the associations of proteins with BMI and BMI changes. In NTR, we investigated the association of gene expression of genes encoding proteins identified in FinnTwin12 with BMI and changes in BMI. We linked identified proteins and their coding genes to plasma metabolites and polygenic risk scores (PRS) using mixed-effect models and correlation networks. Results: We identified 66 and 14 proteins associated with BMI at blood sampling and changes in BMI, respectively. The average heritability of these proteins was 35%. Of the 66 BMI-protein associations, 43 and 12 showed genetic and environmental correlations, respectively, including 8 proteins showing both. Similarly, we observed 6 and 4 genetic and environmental correlations between changes in BMI and protein abundance, respectively. S100A8 gene expression was associated with BMI at blood sampling, and the PRG4 and CFI genes were associated with BMI changes. Proteins showed strong connections with many metabolites and PRSs, but we observed no multi-omics connections among gene expression and other omics layers. Conclusions: Associations between the proteome and BMI trajectories are characterized by shared genetic, environmental, and metabolic etiologies. We observed few gene-protein pairs associated with BMI or changes in BMI at the proteome and transcriptome levels.

Cluster Headache Genomewide Association Study and Meta-Analysis Identifies Eight Loci and Implicates Smoking as Causal Risk Factor.

OBJECTIVE: The objective of this study was to aggregate data for the first genomewide association study meta-analysis of cluster headache, to identify genetic risk variants, and gain biological insights. METHODS: A total of 4,777 cases (3,348 men and 1,429 women) with clinically diagnosed cluster headache were recruited from 10 European and 1 East Asian cohorts. We first performed an inverse-variance genomewide association meta-analysis of 4,043 cases and 21,729 controls of European ancestry. In a secondary trans-ancestry meta-analysis, we included 734 cases and 9,846 controls of East Asian ancestry. Candidate causal genes were prioritized by 5 complementary methods: expression quantitative trait loci, transcriptome-wide association, fine-mapping of causal gene sets, genetically driven DNA methylation, and effects on protein structure. Gene set and tissue enrichment analyses, genetic correlation, genetic risk score analysis, and Mendelian randomization were part of the downstream analyses. RESULTS: The estimated single nucleotide polymorphism (SNP)-based heritability of cluster headache was 14.5%. We identified 9 independent signals in 7 genomewide significant loci in the primary meta-analysis, and one additional locus in the trans-ethnic meta-analysis. Five of the loci were previously known. The 20 genes prioritized as potentially causal for cluster headache showed enrichment to artery and brain tissue. Cluster headache was genetically correlated with cigarette smoking, risk-taking behavior, attention deficit hyperactivity disorder (ADHD), depression, and musculoskeletal pain. Mendelian randomization analysis indicated a causal effect of cigarette smoking intensity on cluster headache. Three of the identified loci were shared with migraine. INTERPRETATION: This first genomewide association study meta-analysis gives clues to the biological basis of cluster headache and indicates that smoking is a causal risk factor. ANN NEUROL 2023;94:713-726.

Estimating Dementia Risk Using Multifactorial Prediction Models.

Importance: The clinical value of current multifactorial algorithms for individualized assessment of dementia risk remains unclear. Objective: To evaluate the clinical value associated with 4 widely used dementia risk scores in estimating 10-year dementia risk. Design, Setting, and Participants: This prospective population-based UK Biobank cohort study assessed 4 dementia risk scores at baseline (2006-2010) and ascertained incident dementia during the following 10 years. Replication with a 20-year follow-up was based on the British Whitehall II study. For both analyses, participants who had no dementia at baseline, had complete data on at least 1 dementia risk score, and were linked to electronic health records from hospitalizations or mortality were included. Data analysis was conducted from July 5, 2022, to April 20, 2023. Exposures: Four existing dementia risk scores: the Cardiovascular Risk Factors, Aging and Dementia (CAIDE)-Clinical score, the CAIDE-APOE-supplemented score, the Brief Dementia Screening Indicator (BDSI), and the Australian National University Alzheimer Disease Risk Index (ANU-ADRI). Main Outcomes and Measures: Dementia was ascertained from linked electronic health records. To evaluate how well each score predicted the 10-year risk of dementia, concordance (C) statistics, detection rate, false-positive rate, and the ratio of true to false positives were calculated for each risk score and for a model including age alone. Results: Of 465 929 UK Biobank participants without dementia at baseline (mean [SD] age, 56.5 [8.1] years; range, 38-73 years; 252 778 [54.3%] female participants), 3421 were diagnosed with dementia at follow-up (7.5 per 10 000 person-years). If the threshold for a positive test result was calibrated to achieve a 5% false-positive rate, all 4 risk scores detected 9% to 16% of incident dementia and therefore missed 84% to 91% (failure rate). The corresponding failure rate was 84% for a model that included age only. For a positive test result calibrated to detect at least half of future incident dementia, the ratio of true to false positives ranged between 1 to 66 (for CAIDE-APOE-supplemented) and 1 to 116 (for ANU-ADRI). For age alone, the ratio was 1 to 43. The C statistic was 0.66 (95% CI, 0.65-0.67) for the CAIDE clinical version, 0.73 (95% CI, 0.72-0.73) for the CAIDE-APOE-supplemented, 0.68 (95% CI, 0.67-0.69) for BDSI, 0.59 (95% CI, 0.58-0.60) for ANU-ADRI, and 0.79 (95% CI, 0.79-0.80) for age alone. Similar C statistics were seen for 20-year dementia risk in the Whitehall II study cohort, which included 4865 participants (mean [SD] age, 54.9 [5.9] years; 1342 [27.6%] female participants). In a subgroup analysis of same-aged participants aged 65 (±1) years, discriminatory capacity of risk scores was low (C statistics between 0.52 and 0.60). Conclusions and Relevance: In these cohort studies, individualized assessments of dementia risk using existing risk prediction scores had high error rates. These findings suggest that the scores were of limited value in targeting people for dementia prevention. Further research is needed to develop more accurate algorithms for estimation of dementia risk.

linemodels: clustering effects based on linear relationships.

SUMMARY: Estimation of effects of multiple explanatory variables on multiple outcome measures has become routine across life sciences with high-throughput molecular technologies. The linemodels R-package allows a probabilistic clustering of variables based on their observed effect sizes on two outcomes. AVAILABILITY AND IMPLEMENTATION: An open source implementation in R available at github.com/mjpirinen/linemodels.

C9orf72 hexanucleotide repeat allele tagging SNPs: Associations with ALS risk and longevity.

C9orf72 hexanucleotide repeat expansion is a common cause of amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). The C9orf72 locus may harbor residual risk outside the hexanucleotide repeat expansion, but the evidence is conflicting. Here, we first compared 683 unrelated amyotrophic lateral sclerosis cases and 3,196 controls with Finnish ancestry to find best single nucleotide polymorphisms that tag the C9orf72 hexanucleotide repeat expansion and intermediate-length alleles. Rs2814707 was the best tagging single nucleotide polymorphisms for intermediate-length alleles with ≥7 repeats (p = 5 × 10-307) and rs139185008 for the hexanucleotide repeat expansion (p = 7 × 10-114) as well as alleles with ≥20 repeats. rs139185008*C associated with amyotrophic lateral sclerosis after removing cases with the hexanucleotide repeat expansion, especially in the subpopulation homozygous for the rs2814707*T (p = 0.0002, OR = 5.06), which supports the concept of residual amyotrophic lateral sclerosis risk at the C9orf72 haplotypes other than the hexanucleotide repeat expansion. We then leveraged Finnish biobank data to test the effects of rs2814707*T and rs139185008*C on longevity after removing individuals with amyotrophic lateral sclerosis / frontotemporal dementia diagnoses. In the discovery cohort (n = 230,006), the frequency of rs139185008*C heterozygotes decreased significantly with age in the comparisons between 50 and 80 years vs. >80 years (p = 0.0005) and <50 years vs. >80 years (p = 0.0001). The findings were similar but less significant in a smaller replication cohort (2-sided p = 0.037 in 50-80 years vs. >80 years and 0.061 in <50 years vs. >80 years). Analysis of the allele frequencies in 5-year bins demonstrated that the decrease of rs139185008*C started after the age of 70 years. The hexanucleotide repeat expansion tagging single nucleotide polymorphisms decreasing frequency with age suggests its' association with age-related diseases probably also outside amyotrophic lateral sclerosis / frontotemporal dementia.

Population analyses of mosaic X chromosome loss identify genetic drivers and widespread signatures of cellular selection.

Mosaic loss of the X chromosome (mLOX) is the most commonly occurring clonal somatic alteration detected in the leukocytes of women, yet little is known about its genetic determinants or phenotypic consequences. To address this, we estimated mLOX in >900,000 women across eight biobanks, identifying 10% of women with detectable X loss in approximately 2% of their leukocytes. Out of 1,253 diseases examined, women with mLOX had an elevated risk of myeloid and lymphoid leukemias and pneumonia. Genetic analyses identified 49 common variants influencing mLOX, implicating genes with established roles in chromosomal missegregation, cancer predisposition, and autoimmune diseases. Complementary exome-sequence analyses identified rare missense variants in FBXO10 which confer a two-fold increased risk of mLOX. A small fraction of these associations were shared with mosaic Y chromosome loss in men, suggesting different biological processes drive the formation and clonal expansion of sex chromosome missegregation events. Allelic shift analyses identified alleles on the X chromosome which are preferentially retained, demonstrating that variation at many loci across the X chromosome is under cellular selection. A novel polygenic score including 44 independent X chromosome allelic shift loci correctly inferred the retained X chromosomes in 80.7% of mLOX cases in the top decile. Collectively our results support a model where germline variants predispose women to acquiring mLOX, with the allelic content of the X chromosome possibly shaping the magnitude of subsequent clonal expansion.

Clinical significance of overall assessment of tumor-infiltrating lymphocytes in oropharyngeal cancer: A meta-analysis.

BACKGROUND: The overall assessment of tumor-infiltrating lymphocytes (TILs) evaluated using hematoxylin and eosin (HE) staining has been recently studied in oropharyngeal squamous cell carcinoma (OPSCC). METHODS: We conducted a systematic review of Scopus, Ovid Medline, PubMed, Web of Science, and Cochrane Library to retrieve studies assessing TILs in HE-stained sections of OPSCC. We used fixed-effect models and random-effect models to estimate the pooled hazard ratios (HRs) and confidence intervals (CIs) for disease-free survival (DFS), overall survival (OS) and disease-specific survival (DSS). RESULTS: Eleven studies were identified that had analyzed the prognostic significance of TILs in OPSCC using HE-stained specimens. Our meta-analyses showed that a high infiltration of TILs was significantly associated with improved DFS (HR 0.39, 95%CI 0.24-0.65, P = 0.0003), OS (HR 0.38, 95%CI 0.29-0.50, P < 0.0001), and DSS (HR 0.32, 95%CI 0.19-0.53, P < 0.0001). CONCLUSION: Findings of our meta-analysis support a growing body of evidence indicating that assessment of TILs in OPSCC using HE-stained sections has reliable prognostic value. The clinical significance of such assessment of TILs has been reported repeatedly in many studies on OPSCC. The assessment is cost-effective, feasible, easy to transfer from lab to clinic, and therefore can be incorporated in daily practice.

Genome-wide screen of otosclerosis in population biobanks: 27 loci and shared associations with skeletal structure.

Otosclerosis is one of the most common causes of conductive hearing loss, affecting 0.3% of the population. It typically presents in adulthood and half of the patients have a positive family history. The pathophysiology of otosclerosis is poorly understood. A previous genome-wide association study (GWAS) identified a single association locus in an intronic region of RELN. Here, we report a meta-analysis of GWAS studies of otosclerosis in three population-based biobanks comprising 3504 cases and 861,198 controls. We identify 23 novel risk loci (p < 5 × 10-8) and report an association in RELN and three previously reported candidate gene or linkage regions (TGFB1, MEPE, and OTSC7). We demonstrate developmental stage-dependent immunostaining patterns of MEPE and RUNX2 in mouse otic capsules. In most association loci, the nearest protein-coding genes are implicated in bone remodelling, mineralization or severe skeletal disorders. We highlight multiple genes involved in transforming growth factor beta signalling for follow-up studies.

Inflammatory and infectious upper respiratory diseases associate with 41 genomic loci and type 2 inflammation.

Inflammatory and infectious upper respiratory diseases (ICD-10: J30-J39), such as diseases of the sinonasal tract, pharynx and larynx, are growing health problems yet their genomic similarity is not known. We analyze genome-wide association to eight upper respiratory diseases (61,195 cases) among 260,405 FinnGen participants, meta-analyzing diseases in four groups based on an underlying genetic correlation structure. Aiming to understand which genetic loci contribute to susceptibility to upper respiratory diseases in general and its subtypes, we detect 41 independent genome-wide significant loci, distinguishing impact on sinonasal or pharyngeal diseases, or both. Fine-mapping implicated non-synonymous variants in nine genes, including three linked to immune-related diseases. Phenome-wide analysis implicated asthma and atopic dermatitis at sinonasal disease loci, and inflammatory bowel diseases and other immune-mediated disorders at pharyngeal disease loci. Upper respiratory diseases also genetically correlated with autoimmune diseases such as rheumatoid arthritis, autoimmune hypothyroidism, and psoriasis. Finally, we associated separate gene pathways in sinonasal and pharyngeal diseases that both contribute to type 2 immunological reaction. We show shared heritability among upper respiratory diseases that extends to several immune-mediated diseases with diverse mechanisms, such as type 2 high inflammation.

Genetic predictors of lifelong medication-use patterns in cardiometabolic diseases.

Little is known about the genetic determinants of medication use in preventing cardiometabolic diseases. Using the Finnish nationwide drug purchase registry with follow-up since 1995, we performed genome-wide association analyses of longitudinal patterns of medication use in hyperlipidemia, hypertension and type 2 diabetes in up to 193,933 individuals (55% women) in the FinnGen study. In meta-analyses of up to 567,671 individuals combining FinnGen with the Estonian Biobank and the UK Biobank, we discovered 333 independent loci (P < 5 × 10-9) associated with medication use. Fine-mapping revealed 494 95% credible sets associated with the total number of medication purchases, changes in medication combinations or treatment discontinuation, including 46 credible sets in 40 loci not associated with the underlying treatment targets. The polygenic risk scores (PRS) for cardiometabolic risk factors were strongly associated with the medication-use behavior. A medication-use enhanced multitrait PRS for coronary artery disease matched the performance of a risk factor-based multitrait coronary artery disease PRS in an independent sample (UK Biobank, n = 343,676). In summary, we demonstrate medication-based strategies for identifying cardiometabolic risk loci and provide genome-wide tools for preventing cardiovascular diseases.

Genetic analyses implicate complex links between adult testosterone levels and health and disease.

BACKGROUND: Testosterone levels are linked with diverse characteristics of human health, yet, whether these associations reflect correlation or causation remains debated. Here, we provide a broad perspective on the role of genetically determined testosterone on complex diseases in both sexes. METHODS: Leveraging genetic and health registry data from the UK Biobank and FinnGen (total N = 625,650), we constructed polygenic scores (PGS) for total testosterone, sex-hormone binding globulin (SHBG) and free testosterone, associating these with 36 endpoints across different disease categories in the FinnGen. These analyses were combined with Mendelian Randomization (MR) and cross-sex PGS analyses to address causality. RESULTS: We show testosterone and SHBG levels are intricately tied to metabolic health, but report lack of causality behind most associations, including type 2 diabetes (T2D). Across other disease domains, including 13 behavioral and neurological diseases, we similarly find little evidence for a substantial contribution from normal variation in testosterone levels. We nonetheless find genetically predicted testosterone affects many sex-specific traits, with a pronounced impact on female reproductive health, including causal contribution to PCOS-related traits like hirsutism and post-menopausal bleeding (PMB). We also illustrate how testosterone levels associate with antagonistic effects on stroke risk and reproductive endpoints between the sexes. CONCLUSIONS: Overall, these findings provide insight into how genetically determined testosterone correlates with several health parameters in both sexes. Yet the lack of evidence for a causal contribution to most traits beyond sex-specific health underscores the complexity of the mechanisms linking testosterone levels to disease risk and sex differences.

Hormones, such as testosterone, travel around the body communicating between the different parts. Testosterone is present at higher levels in men, but also present in women. Variable testosterone levels explain some differences in human traits and disease prevalence. Here, we study how adult testosterone levels relate to health and disease. Genetic, i.e. inherited, differences in testosterone levels contribute to many traits specific to men or women, such as women's reproductive health, hormonal cancers, and hair growth typical in males. However, testosterone levels do not appear as a major cause of most traits studied, including psychiatric diseases and metabolic health. Normal variation in baseline testosterone levels thus seems to have a relatively minor impact on health and disease.

Integration of Biomarker Polygenic Risk Score Improves Prediction of Coronary Heart Disease.

There are several established biomarkers for coronary heart disease (CHD), including blood pressure, cholesterol, and lipoproteins. It is of high interest to determine how a combined polygenic risk score (PRS) of CHD-associated biomarkers (BioPRS) can further improve genetic prediction of CHD. We developed CHDBioPRS, combining BioPRS with PRS of CHD in the UK Biobank and tested it on FinnGen. We found that BioPRS was clearly predictive of CHD and that CHDBioPRS improved the standard CHD PRS. The largest effect was observed with early onset cases in FinnGen, with HRs above 2 per standard deviation of CHDBioPRS.

Lifestyle differences between co-twins are associated with decreased similarity in their internal and external exposome profiles.

Whether differences in lifestyle between co-twins are reflected in differences in their internal or external exposome profiles remains largely underexplored. We therefore investigated whether within-pair differences in lifestyle were associated with within-pair differences in exposome profiles across four domains: the external exposome, proteome, metabolome and epigenetic age acceleration (EAA). For each domain, we assessed the similarity of co-twin profiles using Gaussian similarities in up to 257 young adult same-sex twin pairs (54% monozygotic). We additionally tested whether similarity in one domain translated into greater similarity in another. Results suggest that a lower degree of similarity in co-twins' exposome profiles was associated with greater differences in their behavior and substance use. The strongest association was identified between excessive drinking behavior and the external exposome. Overall, our study demonstrates how social behavior and especially substance use are connected to the internal and external exposomes, while controlling for familial confounders.

Lipidome- and Genome-Wide Study to Understand Sex Differences in Circulatory Lipids.

Background Despite well-recognized differences in the atherosclerotic cardiovascular disease risk between men and women, sex differences in risk factors and sex-specific mechanisms in the pathophysiology of atherosclerotic cardiovascular disease remain poorly understood. Lipid metabolism plays a central role in the development of atherosclerotic cardiovascular disease. Understanding sex differences in lipids and their genetic determinants could provide mechanistic insights into sex differences in atherosclerotic cardiovascular disease and aid in precise risk assessment. Herein, we examined sex differences in plasma lipidome and heterogeneity in genetic influences on lipidome in men and women through sex-stratified genome-wide association analyses. Methods and Results We used data consisting of 179 lipid species measured by shotgun lipidomics in 7266 individuals from the Finnish GeneRISK cohort and sought for replication using independent data from 2045 participants. Significant sex differences in the levels of 141 lipid species were observed (P<7.0×10-4). Interestingly, 121 lipid species showed significant age-sex interactions, with opposite age-related changes in 39 lipid species. In general, most of the cholesteryl esters, ceramides, lysophospholipids, and glycerides were higher in 45- to 50-year-old men compared with women of same age, but the sex differences narrowed down or reversed with age. We did not observe any major differences in genetic effect in the sex-stratified genome-wide association analyses, which suggests that common genetic variants do not have a major role in sex differences in lipidome. Conclusions Our study provides a comprehensive view of sex differences in circulatory lipids pointing to potential sex differences in lipid metabolism and highlights the need for sex- and age-specific prevention strategies.