Polygenic prediction via Bayesian regression and continuous shrinkage priors

The Associations of Polygenic Scores for Risky Behaviors and Parenting Behaviors with Adolescent Externalizing Problems

The current study focused on longitudinal effects of genetics and parental behaviors and their interplay on externalizing behaviors in a panel study following individuals from adolescence to young adulthood. The nationally representative sample of Add Health participants of European ancestry included N = 4142 individuals, measured on three occasions. Parenting was operationalized as experiences with child maltreatment and maternal closeness. Externalizing problems were operationalized as alcohol use, cannabis use, and antisocial behaviors. Genetic effects were operationalized as a polygenic score (PGS) of risky behaviors. The results showed significant effects for child maltreatment, maternal closeness, and PGS, above and beyond other factors and previous levels of externalizing behaviors. Furthermore, maternal closeness was found to negatively correlate with PGS. No significant interaction effects of parenting and PGS were found. The results underscore the joint independent effects of parenting and genetics on the change in externalizing behaviors from adolescence to young adulthood.

Towards Predictive Analysis on Disease Progression: A Variational Hawkes Process Model

Massively available longitudinal data about long-term disease trajectories of patients provides a golden mine for the understanding of disease progression and efficient health service delivery. It calls for quantitative modeling of disease progression, which is a tricky problem due to the complexity of the disease progression process as well as the irregularity of time documented in trajectories. In this study, we tackle the problem with the goal of predictively analyzing disease progression. Specifically, we propose a novel Variational Hawkes Process (VHP) model to generalize disease progression and predict future patient states based on the clinical observational data of past disease trajectories. First, Hawkes Process captures the intensity of irregular visits in a trajectory documented on medical facilities and controls the aforementioned information flowing into future visits. Thereafter, the captured intensity is incorporated into a Variational Auto-Encoder to generate the representation of the future partial disease trajectory for a target patient in a predictive manner. To further improve the prediction performance, we equip the proposed model with a disease trajectory discriminator to distinguish the generated trajectories from real ones. We evaluate the proposed model on two public datasets from the MIMIC-III database pertaining to heart failure and sepsis patients, respectively, and one real-world dataset from a Chinese hospital pertaining to heart failure patients with multiple admissions. Experimental results demonstrate that the proposed model significantly outperforms state-of-the-art baselines, and may derive a set of practical implications that can benefit a wide spectrum of management and applications on disease progression.

A fast and robust Bayesian nonparametric method for prediction of complex traits using summary statistics

Genetic prediction of complex traits has great promise for disease prevention, monitoring, and treatment. The development of accurate risk prediction models is hindered by the wide diversity of genetic architecture across different traits, limited access to individual level data for training and parameter tuning, and the demand for computational resources. To overcome the limitations of the most existing methods that make explicit assumptions on the underlying genetic architecture and need a separate validation data set for parameter tuning, we develop a summary statistics-based nonparametric method that does not rely on validation datasets to tune parameters. In our implementation, we refine the commonly used likelihood assumption to deal with the discrepancy between summary statistics and external reference panel. We also leverage the block structure of the reference linkage disequilibrium matrix for implementation of a parallel algorithm. Through simulations and applications to twelve traits, we show that our method is adaptive to different genetic architectures, statistically robust, and computationally efficient. Our method is available at https://github.com/eldronzhou/SDPR.

Identifying individuals with high risk of Alzheimer's disease using polygenic risk scores

Polygenic Risk Scores (PRS) for AD offer unique possibilities for reliable identification of individuals at high and low risk of AD. However, there is little agreement in the field as to what approach should be used for genetic risk score calculations, how to model the effect of APOE, what the optimal p-value threshold (pT) for SNP selection is and how to compare scores between studies and methods. We show that the best prediction accuracy is achieved with a model with two predictors (APOE and PRS excluding APOE region) with pT<0.1 for SNP selection. Prediction accuracy in a sample across different PRS approaches is similar, but individuals' scores and their associated ranking differ. We show that standardising PRS against the population mean, as opposed to the sample mean, makes the individuals' scores comparable between studies. Our work highlights the best strategies for polygenic profiling when assessing individuals for AD risk.

Optimal linkage disequilibrium splitting

MOTIVATION

A few algorithms have been developed for splitting the genome in nearly independent blocks of linkage disequilibrium. Due to the complexity of this problem, these algorithms rely on heuristics, which makes them suboptimal.

RESULTS

Here, we develop an optimal solution for this problem using dynamic programming.

AVAILABILITY

This is now implemented as function snp_ldsplit as part of R package bigsnpr.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

A novel transcriptional risk score for risk prediction of complex human diseases

Recently polygenetic risk score (PRS) has been successfully used in the risk prediction of complex human diseases. Many studies incorporated internal information, such as effect size distribution, or external information, such as linkage disequilibrium, functional annotation, and pleiotropy among multiple diseases, to optimize the performance of PRS. To leverage on multiomics datasets, we developed a novel flexible transcriptional risk score (TRS), in which messenger RNA expression levels were imputed and weighted for risk prediction. In simulation studies, we demonstrated that single-tissue TRS has greater prediction power than LDpred, especially when there is a large effect of gene expression on the phenotype. Multitissue TRS improves prediction accuracy when there are multiple tissues with independent contributions to disease risk. We applied our method to complex traits, including Crohn's disease, type 2 diabetes, and so on. The single-tissue TRS method outperformed LDpred and AnnoPred across the tested traits. The performance of multitissue TRS is trait-dependent. Moreover, our method can easily incorporate information from epigenomic and proteomic data upon the availability of reference datasets.

Clinical Conditions and Their Impact on Utility of Genetic Scores for Prediction of Acute Coronary Syndrome

BACKGROUND

Acute coronary syndrome (ACS) is a clinically significant presentation of coronary heart disease. Genetic information has been proposed to improve prediction beyond well-established clinical risk factors. While polygenic scores (PS) can capture an individual's genetic risk for ACS, its prediction performance may vary in the context of diverse correlated clinical conditions. Here, we aimed to test whether clinical conditions impact the association between PS and ACS.

METHODS

We explored the association between 405 clinical conditions diagnosed before baseline and 9080 incident cases of ACS in 387 832 individuals from the UK Biobank. Results were replicated in 6430 incident cases of ACS in 177 876 individuals from FinnGen.

RESULTS

We identified 80 conventional (eg, stable angina pectoris and type 2 diabetes) and unconventional (eg, diaphragmatic hernia and inguinal hernia) associations with ACS. The association between PS and ACS was consistent in individuals with and without most clinical conditions. However, a diagnosis of stable angina pectoris yielded a differential association between PS and ACS. PS was associated with a significantly reduced (interaction P=2.87×10^-8) risk for ACS in individuals with stable angina pectoris (hazard ratio, 1.163 [95% CI, 1.082-1.251]) compared with individuals without stable angina pectoris (hazard ratio, 1.531 [95% CI, 1.497-1.565]). These findings were replicated in FinnGen (interaction P=1.38×10^-6).

CONCLUSIONS

In summary, while most clinical conditions did not impact utility of PS for prediction of ACS, we found that PS was substantially less predictive of ACS in individuals with prevalent stable coronary heart disease. PS may be more appropriate for prediction of ACS in asymptomatic individuals than symptomatic individuals with clinical suspicion for coronary heart disease.

Polygenic risk for immuno-metabolic markers and specific depressive symptoms: A multi-sample network analysis study

BACKGROUND

About every fourth patient with major depressive disorder (MDD) shows evidence of systemic inflammation. Previous studies have shown inflammation-depression associations of multiple serum inflammatory markers and multiple specific depressive symptoms. It remains unclear, however, if these associations extend to genetic/lifetime predisposition to higher inflammatory marker levels and what role metabolic factors such as Body Mass Index (BMI) play. It is also unclear whether inflammation-symptom associations reflect direct or indirect associations, which can be disentangled using network analysis.

METHODS

This study examined associations of polygenic risk scores (PRSs) for immuno-metabolic markers (C-reactive protein [CRP], interleukin [IL]-6, IL-10, tumour necrosis factor [TNF]-α, BMI) with seven depressive symptoms in one general population sample, the UK Biobank study (n = 110,010), and two patient samples, the Munich Antidepressant Response Signature (MARS, n = 1058) and Sequenced Treatment Alternatives to Relieve Depression (STAR*D, n = 1143) studies. Network analysis was applied jointly for these samples using fused graphical least absolute shrinkage and selection operator (FGL) estimation as primary analysis and, individually, using unregularized model search estimation. Stability of results was assessed using bootstrapping and three consistency criteria were defined to appraise robustness and replicability of results across estimation methods, network bootstrapping, and samples.

RESULTS

Network analysis results displayed to-be-expected PRS-PRS and symptom-symptom associations (termed edges), respectively, that were mostly positive. Using FGL estimation, results further suggested 28, 29, and six PRS-symptom edges in MARS, STAR*D, and UK Biobank samples, respectively. Unregularized model search estimation suggested three PRS-symptom edges in the UK Biobank sample. Applying our consistency criteria to these associations indicated that only the association of higher CRP PRS with greater changes in appetite fulfilled all three criteria. Four additional associations fulfilled at least two consistency criteria; specifically, higher CRP PRS was associated with greater fatigue and reduced anhedonia, higher TNF-α PRS was associated with greater fatigue, and higher BMI PRS with greater changes in appetite and anhedonia. Associations of the BMI PRS with anhedonia, however, showed an inconsistent valence across estimation methods.

CONCLUSIONS

Genetic predisposition to higher systemic inflammatory markers are primarily associated with somatic/neurovegetative symptoms of depression such as changes in appetite and fatigue, consistent with previous studies based on circulating levels of inflammatory markers. We extend these findings by providing evidence that associations are direct (using network analysis) and extend to genetic predisposition to immuno-metabolic markers (using PRSs). Our findings can inform selection of patients with inflammation-related symptoms into clinical trials of immune-modulating drugs for MDD.

The associations between polygenic risk, sensation seeking, social support, and alcohol use in adulthood

Genetic predispositions play an important role in alcohol use. Understanding the psychosocial mechanisms through which genetic risk unfolds to influence alcohol use outcomes is critical for identifying modifiable targets and developing prevention and intervention efforts. In this study, we examined the role of sensation seeking and social support from family and friends in linking genetic risk to alcohol use. We also examined the role of social support in moderating the associations between genetic risk and sensation seeking and alcohol use. Data were drawn from a sample of 2,836 European American adults from the Collaborative Study on the Genetics of Alcoholism (46% male, mean age = 35.65, standard deviation [SD] = 10.78). Results from path analysis indicated that genome-wide polygenic scores for alcohol consumption (alc-GPS) were associated with higher sensation seeking, which in turn was associated with higher levels of alcohol use. alc-GPS was also associated with higher alcohol use indirectly via lower levels of family support. In addition, high friend support attenuated the association between alc-GPS and sensation seeking and alcohol use. The pattern of associations was similar for males and females, with some differences in the associations between social support and alcohol use observed across age. Our findings highlight the important role of intermediate phenotypes and gene-environment interplay in the pathways of risk from genetic predispositions to complex alcohol use outcomes. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

Fast Numerical Optimization for Genome Sequencing Data in Population Biobanks

MOTIVATION

Large-scale and high-dimensional genome sequencing data poses computational challenges. General purpose optimization tools are usually not optimal in terms of computational and memory performance for genetic data.

RESULTS

We develop two efficient solvers for optimization problems arising from large-scale regularized regressions on millions of genetic variants sequenced from hundreds of thousands of individuals. These genetic variants are encoded by the values in the set {0, 1, 2, NA}. We take advantage of this fact and use two bits to represent each entry in a genetic matrix, which reduces memory requirement by a factor of 32 compared to a double precision floating point representation. Using this representation, we implemented an iteratively reweighted least square algorithm to solve Lasso regressions on genetic matrices, which we name snpnet-2.0. When the dataset contains many rare variants, the predictors can be encoded in a sparse matrix. We utilize the sparsity in the predictor matrix to further reduce memory requirement and computational speed. Our sparse genetic matrix implementation uses both the compact 2-bit representation and a simplified version of compressed sparse block format so that matrix-vector multiplications can be effectively parallelized on multiple CPU cores. To demonstrate the effectiveness of this representation, we implement an accelerated proximal gradient method to solve group Lasso on these sparse genetic matrices. This solver is named sparse-snpnet, and will also be included as part of snpnet R package. Our implementation is able to solve Lasso and group Lasso, linear, logistic and Cox regression problems on sparse genetic matrices that contain 1,000,000 variants and almost 100,000 individuals within 10 minutes and using less than 32GB of memory.

AVAILABILITY

https://github.com/rivas-lab/snpnet/tree/compact.

RápidoPGS: A rapid polygenic score calculator for summary GWAS data without a test dataset

Motivation

Polygenic scores (PGS) aim to genetically predict complex traits at an individual level. PGS are typically trained on genome-wide association summary statistics and require an independent test dataset to tune parameters. More recent methods allow parameters to be tuned on the training data, removing the need for independent test data, but approaches are computationally intensive. Based on fine-mapping principles, we present RápidoPGS, a flexible and fast method to compute PGS requiring summary-level Genome-wide association studies (GWAS) datasets only, with little computational requirements and no test data required for parameter tuning.

Results

We show that RápidoPGS performs slightly less well than two out of three other widely used PGS methods (LDpred2, PRScs and SBayesR) for case–control datasets, with median r² difference: -0.0092, -0.0042 and 0.0064, respectively, but up to 17 000-fold faster with reduced computational requirements. RápidoPGS is implemented in R and can work with user-supplied summary statistics or download them from the GWAS catalog.

Availability and implementation

Our method is available with a GPL license as an R package from CRAN and GitHub.

Supplementary information

Supplementary data are available at Bioinformatics online.

Polygenic Risk Scores

As genome-wide association studies have continued to identify loci associated with complex traits, the implications of and necessity for proper use of these findings, including prediction of disease risk, have become apparent. Many complex diseases have numerous associated loci with detectable effects implicating risk for or protection from disease. A common contemporary approach to using this information for disease prediction is through the application of genetic risk scores. These scores estimate an individual's liability for a specific outcome by aggregating the effects of associated loci into a single measure as described in the previous version of this article. Although genetic risk scores have traditionally included variants that meet criteria for genome-wide significance, an extension known as the polygenic risk score has been developed to include the effects of more variants across the entire genome. Here, we describe common methods and software packages for calculating and interpreting polygenic risk scores. In this revised version of the article, we detail information that is needed to perform a polygenic risk score analysis, considerations for planning the analysis and interpreting results, as well as discussion of the limitations based on the choices made. We also provide simulated sample data and a walkthrough for four different polygenic risk score software. © 2021 Wiley Periodicals LLC.

Polygenic risk scores for alcohol involvement relate to brain structure in substance-naïve children: Results from the ABCD study

Brain imaging-derived structural correlates of alcohol involvement have largely been speculated to arise as a consequence of alcohol exposure. However, they may also reflect predispositional risk. In substance naïve children of European ancestry who completed the baseline session of the Adolescent Brain Cognitive Development (ABCD) Study (n = 3013), mixed-effects models estimated whether polygenic risk scores (PRS) for problematic alcohol use (PAU-PRS) and drinks per week (DPW-PRS) are associated with magnetic resonance imaging-derived brain structure phenotypes (i.e., total and regional: cortical thickness, surface area and volume; subcortical volume; white matter volume, fractional anisotropy, mean diffusivity). Follow-up analyses evaluated whether any identified regions were also associated with polygenic risk among substance naïve children of African ancestry (n = 898). After adjustment for multiple testing correction, polygenic risk for PAU was associated with lower volume of the left frontal pole and greater cortical thickness of the right supramarginal gyrus (|βs| > 0.009; ps < 0.001; psfdr  < 0.046; r2 s < 0.004). PAU PRS and DPW PRS showed nominally significant associations with a host of other regional brain structure phenotypes (e.g., insula surface area and volume). None of these regions showed any, even nominal association among children of African ancestry. Genomic liability to alcohol involvement may manifest as variability in brain structure during middle childhood prior to alcohol use initiation. Broadly, alcohol-related variability in brain morphometry may partially reflect predisposing genomic influence. Larger discovery genome-wide association studies and target samples of diverse ancestries are needed to determine whether observed associations may generalize across ancestral origins.

Leveraging both individual-level genetic data and GWAS summary statistics increases polygenic prediction

The accuracy of polygenic risk scores (PRSs) to predict complex diseases increases with the training sample size. PRSs are generally derived based on summary statistics from large meta-analyses of multiple genome-wide association studies (GWASs). However, it is now common for researchers to have access to large individual-level data as well, such as the UK Biobank data. To the best of our knowledge, it has not yet been explored how best to combine both types of data (summary statistics and individual-level data) to optimize polygenic prediction. The most widely used approach to combine data is the meta-analysis of GWAS summary statistics (meta-GWAS), but we show that it does not always provide the most accurate PRS. Through simulations and using 12 real case-control and quantitative traits from both iPSYCH and UK Biobank along with external GWAS summary statistics, we compare meta-GWAS with two alternative data-combining approaches, stacked clumping and thresholding (SCT) and meta-PRS. We find that, when large individual-level data are available, the linear combination of PRSs (meta-PRS) is both a simple alternative to meta-GWAS and often more accurate.

Translating genetic association of lipid levels for biological and clinical application

PURPOSE OF REVIEW

This review focuses on the foundational evidence from the last two decades of lipid genetics research and describes the current status of data-driven approaches for transethnic GWAS, fine-mapping, transcriptome informed fine-mapping, and disease prediction.

RECENT FINDINGS

Current lipid genetics research aims to understand the association mechanisms and clinical relevance of lipid loci as well as to capture population specific associations found in global ancestries. Recent genome-wide trans-ethnic association meta-analyses have identified 118 novel lipid loci reaching genome-wide significance. Gene-based burden tests of whole exome sequencing data have identified three genes-PCSK9, LDLR, and APOB-with significant rare variant burden associated with familial dyslipidemia. Transcriptome-wide association studies discovered five previously unreported lipid-associated loci. Additionally, the predictive power of genome-wide genetic risk scores amalgamating the polygenic determinants of lipid levels can potentially be used to increase the accuracy of coronary artery disease prediction.

CONCLUSIONS

Lipids are one of the most successful group of traits in the era of genome-wide genetic discovery for identification of novel loci and plausible drug targets. However, a substantial fraction of lipid trait heritability remains unexplained. Further analysis of diverse ancestries and state of the art methods for association locus refinement could potentially reveal some of this missing heritability and increase the clinical application of the genomic association results.

The trans-ancestral genomic architecture of glycemic traits

Glycemic traits are used to diagnose and monitor type 2 diabetes and cardiometabolic health. To date, most genetic studies of glycemic traits have focused on individuals of European ancestry. Here we aggregated genome-wide association studies comprising up to 281,416 individuals without diabetes (30% non-European ancestry) for whom fasting glucose, 2-h glucose after an oral glucose challenge, glycated hemoglobin and fasting insulin data were available. Trans-ancestry and single-ancestry meta-analyses identified 242 loci (99 novel; P < 5 × 10-8), 80% of which had no significant evidence of between-ancestry heterogeneity. Analyses restricted to individuals of European ancestry with equivalent sample size would have led to 24 fewer new loci. Compared with single-ancestry analyses, equivalent-sized trans-ancestry fine-mapping reduced the number of estimated variants in 99% credible sets by a median of 37.5%. Genomic-feature, gene-expression and gene-set analyses revealed distinct biological signatures for each trait, highlighting different underlying biological pathways. Our results increase our understanding of diabetes pathophysiology by using trans-ancestry studies for improved power and resolution.

Polygenic Risk Score-Enhanced Risk Stratification of Coronary Artery Disease in Patients With Stable Chest Pain

BACKGROUND

Polygenic risk scores (PRSs) are associated with coronary artery disease (CAD), but the clinical potential of using PRSs at the single-patient level for risk stratification has yet to be established. We investigated whether adding a PRS to clinical risk factors (CRFs) improves risk stratification in patients referred to coronary computed tomography angiography on a suspicion of obstructive CAD.

METHODS

In this prespecified diagnostic substudy of the Dan-NICAD trial (Danish study of Non-Invasive testing in Coronary Artery Disease), we included 1617 consecutive patients with stable chest symptoms and no history of CAD referred for coronary computed tomography angiography. CRFs used for risk stratification were age, sex, symptoms, prior or active smoking, antihypertensive treatment, lipid-lowering treatment, and diabetes. In addition, patients were genotyped, and their PRSs were calculated. All patients underwent coronary computed tomography angiography. Patients with a suspected ≥50% stenosis also underwent invasive coronary angiography with fractional flow reserve. A combined end point of obstructive CAD was defined as a visual invasive coronary angiography stenosis >90%, fractional flow reserve <0.80, or a quantitative coronary analysis stenosis >50% if fractional flow reserve measurements were not feasible.

RESULTS

The PRS was associated with obstructive CAD independent of CRFs (adjusted odds ratio, 1.8 [95% CI, 1.5-2.2] per SD). The PRS had an area under the curve of 0.63 (0.59-0.68), which was similar to that for age and sex. Combining the PRS with CRFs led to a CRF+PRS model with area under the curve of 0.75 (0.71-0.79), which was 0.04 more than the CRF model (P=0.0029). By using pretest probability (pretest probability) cutoffs at 5% and 15%, a net reclassification improvement of 15.8% (P=3.1×10^-4) was obtained, with a down-classification of risk in 24% of patients (211 of 862) in whom the pretest probability was 5% to 15% based on CRFs alone.

CONCLUSIONS

Adding a PRS improved risk stratification of obstructive CAD beyond CRFs, suggesting a modest clinical potential of using PRSs to guide diagnostic testing in the contemporary clinical setting. Registration: URL: https://www.clinicaltrials.gov; Unique identifier: NCT02264717.

Imputed gene expression risk scores: a functionally informed component of polygenic risk

Integration of functional genomic annotations when estimating polygenic risk scores (PRS) can provide insight into aetiology and improve risk prediction. This study explores the predictive utility of gene expression risk scores (GeRS), calculated using imputed gene expression and transcriptome-wide association study (TWAS) results. The predictive utility of GeRS was evaluated using 12 neuropsychiatric and anthropometric outcomes measured in two target samples: UK Biobank and the Twins Early Development Study. GeRS were calculated based on imputed gene expression levels and TWAS results, using 53 gene expression-genotype panels, termed single nucleotide polymorphism (SNP)-weight sets, capturing expression across a range of tissues. We compare the predictive utility of elastic net models containing GeRS within and across SNP-weight sets, and models containing both GeRS and PRS. We estimate the proportion of SNP-based heritability attributable to cis-regulated gene expression. GeRS significantly predicted a range of outcomes, with elastic net models combining GeRS across SNP-weight sets improving prediction. GeRS were less predictive than PRS, but models combining GeRS and PRS improved prediction for several outcomes, with relative improvements ranging from 0.3% for height (P = 0.023) to 4% for rheumatoid arthritis (P = 5.9 × 10-8). The proportion of SNP-based heritability attributable to cis-regulated expression was modest for most outcomes, even when restricting GeRS to colocalized genes. GeRS represent a component of PRS and could be useful for functional stratification of genetic risk. Only in specific circumstances can GeRS substantially improve prediction over PRS alone. Future research considering functional genomic annotations when estimating genetic risk is warranted.

Association between the polygenic liabilities for prostate cancer and breast cancer with biochemical recurrence after radical prostatectomy for localized prostate cancer

Prostate and breast cancers are hormone-related malignancies and are characterized by a complex interplay of hundreds of susceptibility loci throughout the genome. Prostate cancer could be inhibited by eliminating androgens through castration or estrogen administration, thus facilitating long-term treatment of prostate cancer; however, the role of estrogen in prostate cancer remains unclear. This study aimed to determine whether polygenic risk scores (PRSs) comprising combinations of genome-wide susceptibility variants influence the clinical outcomes of prostate cancer patients. The study subjects were recruited from four medical centers in Taiwan, and genome-wide genotyping data were obtained from 643 prostate cancer patients. We derived the PRS for prostate cancer (PRS-PC) and for breast cancer (PRS-BC) for each patient. The association between the PRS-PC/PRS-BC at the age of prostate cancer onset and recurrence within seven years was evaluated using a regression model adjusted for population stratification components. A higher PRS-PC was associated with an earlier onset age for prostate cancer (beta in per SD increase in PRS = -0.89, P = 0.0008). In contrast, a higher PRS-BC was associated with an older onset age for prostate cancer (beta = 0.59, P = 0.02). PRS-PC was not associated with the risk of recurrence (hazard ratio = 1.03, P = 0.67), whereas a higher PRS-BC was associated with a low recurrence risk (hazard ratio = 0.86, P = 0.03). These results indicate that the genetic predisposition to breast cancer is associated with a low risk of prostate cancer recurrence. Further studies are warranted to explore the role of breast cancer susceptibility variants and estrogen signaling in prostate cancer progression.

Fate or coincidence: do COPD and major depression share genetic risk factors?

Major depressive disorder (MDD) is a common comorbidity in chronic obstructive pulmonary disease (COPD), affecting up to 57% of patients with COPD. Although the comorbidity of COPD and MDD is well established, the causal relationship between these two diseases is unclear. A large-scale electronic health record clinical biobank and genome-wide association study summary statistics for MDD and lung function traits were used to investigate potential shared underlying genetic susceptibility between COPD and MDD. Linkage disequilibrium score regression was used to estimate genetic correlation between phenotypes. Polygenic risk scores (PRS) for MDD and lung function traits were developed and used to perform a phenome-wide association study (PheWAS). Multi-trait-based conditional and joint analysis identified single-nucleotide polymorphisms (SNPs) influencing both lung function and MDD. We found genetic correlations between MDD and all lung function traits were small and not statistically significant. A PRS-MDD was significantly associated with an increased risk of COPD in a PheWAS [odds ratio (OR) = 1.12, 95% confidence interval (CI): 1.09-1.16] when adjusting for age, sex and genetic ancestry, but this relationship became attenuated when controlling for smoking history (OR = 1.08, 95% CI: 1.04-1.13). No significant associations were found between the lung function PRS and MDD. Multi-trait-based conditional and joint analysis identified three SNPs that may contribute to both traits, two of which were previously associated with mood disorders and COPD. Our findings suggest that the observed relationship between COPD and MDD may not be driven by a strong shared genetic architecture.

Associations between Inhaled Corticosteroid in the First 6 Years of Life and Obesity Related Traits

Rationale Infants and young children might be particularly susceptible to the potential clinical side effects of inhaled corticosteroids (ICS) on body mass index (BMI), adiposity rebound and body composition, but this has rarely been studied in long-term studies in this age-group. Objective To determine the association between ICS exposure in the first 6 years of life on BMI, adiposity rebound, body composition and blood lipid levels. Methods Children from the two Copenhagen Prospective Studies on Asthma in Childhood mother-child cohorts were included. ICS use was registered prospectively to age 6 and the cumulative dose was calculated. Multiple linear regression models were used for analysis. Measurements and Main Results A total of 932 (84%) of the 1111 children from the COPSAC cohorts had data on BMI, 786 (71%) had DXA scan data at age 6 years, and 815 (73%) had adiposity rebound age calculated. 291 children (31%) received a cumulative ICS dose > 10 weeks of standard treatment before age 6. ICS treatment during 0-6 years of age was associated with an increased BMI z-score; 0.05 SD [95% CI: 0.005 to 0.09] per one-year standard treatment, p=0.03, an earlier age at adiposity rebound; -0.18 year [95% CI: -0.28 to -0.08], p=0.0006, and a 2 % increased geometric mean android fat percentage, p=0.05. ICS exposure and DXA scan data were not associated. Conclusions ICS use in early childhood was associated with increased BMI z-score at age six, an earlier adiposity rebound and a trend of association with increased android body fat percentage.

Genetic analysis of obstructive sleep apnoea discovers a strong association with cardiometabolic health

There is currently limited understanding of the genetic aetiology of obstructive sleep apnoea (OSA). We aimed to identify genetic loci associated with OSA risk, and to test if OSA and its comorbidities share a common genetic background.We conducted the first large-scale genome-wide association study of OSA using the FinnGen study (217 955 individuals) with 16 761 OSA patients identified using nationwide health registries.We estimated 0.08 (95% CI 0.06-0.11) heritability and identified five loci associated with OSA (p<5.0×10^-8): rs4837016 near GAPVD1 (GTPase activating protein and VPS9 domains 1), rs10928560 near CXCR4 (C-X-C motif chemokine receptor type 4), rs185932673 near CAMK1D (calcium/calmodulin-dependent protein kinase ID) and rs9937053 near FTO (fat mass and obesity-associated protein; a variant previously associated with body mass index (BMI)). In a BMI-adjusted analysis, an association was observed for rs10507084 near RMST/NEDD1 (rhabdomyosarcoma 2 associated transcript/NEDD1 γ-tubulin ring complex targeting factor). We found high genetic correlations between OSA and BMI (r_g=0.72 (95% CI 0.62-0.83)), and with comorbidities including hypertension, type 2 diabetes, coronary heart disease, stroke, depression, hypothyroidism, asthma and inflammatory rheumatic disease (r_g>0.30). The polygenic risk score for BMI showed 1.98-fold increased OSA risk between the highest and the lowest quintile, and Mendelian randomisation supported a causal relationship between BMI and OSA.Our findings support the causal link between obesity and OSA, and the joint genetic basis between OSA and comorbidities.

Evaluation of polygenic prediction methodology within a reference-standardized framework

The predictive utility of polygenic scores is increasing, and many polygenic scoring methods are available, but it is unclear which method performs best. This study evaluates the predictive utility of polygenic scoring methods within a reference-standardized framework, which uses a common set of variants and reference-based estimates of linkage disequilibrium and allele frequencies to construct scores. Eight polygenic score methods were tested: p-value thresholding and clumping (pT+clump), SBLUP, lassosum, LDpred1, LDpred2, PRScs, DBSLMM and SBayesR, evaluating their performance to predict outcomes in UK Biobank and the Twins Early Development Study (TEDS). Strategies to identify optimal p-value thresholds and shrinkage parameters were compared, including 10-fold cross validation, pseudovalidation and infinitesimal models (with no validation sample), and multi-polygenic score elastic net models. LDpred2, lassosum and PRScs performed strongly using 10-fold cross-validation to identify the most predictive p-value threshold or shrinkage parameter, giving a relative improvement of 16-18% over pT+clump in the correlation between observed and predicted outcome values. Using pseudovalidation, the best methods were PRScs, DBSLMM and SBayesR. PRScs pseudovalidation was only 3% worse than the best polygenic score identified by 10-fold cross validation. Elastic net models containing polygenic scores based on a range of parameters consistently improved prediction over any single polygenic score. Within a reference-standardized framework, the best polygenic prediction was achieved using LDpred2, lassosum and PRScs, modeling multiple polygenic scores derived using multiple parameters. This study will help researchers performing polygenic score studies to select the most powerful and predictive analysis methods.

Alzheimer's disease genetic risk and sleep phenotypes in healthy young men: association with more slow waves and daytime sleepiness

STUDY OBJECTIVES

Sleep disturbances and genetic variants have been identified as risk factors for Alzheimer's disease (AD). Our goal was to assess whether genome-wide polygenic risk scores (PRS) for AD associate with sleep phenotypes in young adults, decades before typical AD symptom onset.

METHODS

We computed whole-genome PRS for AD and extensively phenotyped sleep under different sleep conditions, including baseline sleep, recovery sleep following sleep deprivation, and extended sleep opportunity, in a carefully selected homogenous sample of 363 healthy young men (22.1 years ± 2.7) devoid of sleep and cognitive disorders.

RESULTS

AD PRS was associated with more slow-wave energy, that is, the cumulated power in the 0.5-4 Hz EEG band, a marker of sleep need, during habitual sleep and following sleep loss, and potentially with larger slow-wave sleep rebound following sleep deprivation. Furthermore, higher AD PRS was correlated with higher habitual daytime sleepiness.

CONCLUSIONS

These results imply that sleep features may be associated with AD liability in young adults, when current AD biomarkers are typically negative, and support the notion that quantifying sleep alterations may be useful in assessing the risk for developing AD.

Interaction of developmental factors and ordinary stressful life events on brain structure in adults

An interplay of early environmental and genetic risk factors with recent stressful life events (SLEs) in adulthood increases the risk for adverse mental health outcomes. The interaction of early risk and current SLEs on brain structure has hardly been investigated. Whole brain voxel-based morphometry analysis was performed in N = 786 (64.6% female, mean age = 33.39) healthy subjects to identify correlations of brain clusters with commonplace recent SLEs. Genetic and early environmental risk factors, operationalized as those for severe psychopathology (i.e., polygenic scores for neuroticism, childhood maltreatment, urban upbringing and paternal age) were assessed as modulators of the impact of SLEs on the brain. SLEs were negatively correlated with grey matter volume in the left medial orbitofrontal cortex (mOFC, FWE p = 0.003). This association was present for both, positive and negative, life events. Cognitive-emotional variables, i.e., neuroticism, perceived stress, trait anxiety, intelligence, and current depressive symptoms did not account for the SLE-mOFC association. Further, genetic and environmental risk factors were not correlated with grey matter volume in the left mOFC cluster and did not affect the association between SLEs and left mOFC grey matter volume. The orbitofrontal cortex has been implicated in stress-related psychopathology, particularly major depression in previous studies. We find that SLEs are associated with this area. Important early life risk factors do not interact with current SLEs on brain morphology in healthy subjects.

Genomic Prediction Using Bayesian Regression Models With Global-Local Prior

Bayesian regression models are widely used in genomic prediction for various species. By introducing the global parameter τ, which can shrink marker effects to zero, and the local parameter λ_k, which can allow markers with large effects to escape from the shrinkage, we developed two novel Bayesian models, named BayesHP and BayesHE. The BayesHP model uses Horseshoe+ prior, whereas the BayesHE model assumes local parameter λ_k, after a half-t distribution with an unknown degree of freedom. The performances of BayesHP and BayesHE models were compared with three classical prediction models, including GBLUP, BayesA, and BayesB, and BayesU, which also applied global–local prior (Horseshoe prior). To assess model performances for traits with various genetic architectures, simulated data and real data in cattle (milk production, health, and type traits) and mice (type and growth traits) were analyzed. The results of simulation data analysis indicated that models based on global–local priors, including BayesU, BayesHP, and BayesHE, performed better in traits with higher heritability and fewer quantitative trait locus. The results of real data analysis showed that BayesHE was optimal or suboptimal for all traits, whereas BayesHP was not superior to other classical models. For BayesHE, its flexibility to estimate hyperparameter automatically allows the model to be more adaptable to a wider range of traits. The BayesHP model, however, tended to be suitable for traits having major/large quantitative trait locus, given its nature of the “U” type-like shrinkage pattern. Our results suggested that auto-estimate the degree of freedom (e.g., BayesHE) would be a better choice other than increasing the local parameter layers (e.g., BayesHP). In this study, we introduced the global–local prior with unknown hyperparameter to Bayesian regression models for genomic prediction, which can trigger further investigations on model development.

Alcohol and cigarette smoking consumption as genetic proxies for alcohol misuse and nicotine dependence

PURPOSE

To investigate the role of consumption phenotypes as genetic proxies for alcohol misuse and nicotine dependence.

METHODS

We leveraged GWAS data from well-powered studies of consumption, alcohol misuse, and nicotine dependence phenotypes measured in individuals of European ancestry from the UK Biobank (UKB) and other population-based cohorts (largest total N = 263,954), and performed genetic correlations within a medical-center cohort, BioVU (N = 66,915). For alcohol, we used quantitative measures of consumption and misuse via AUDIT from UKB. For smoking, we used cigarettes per day from UKB and non-UKB cohorts comprising the GSCAN consortium, and nicotine dependence via ICD codes from UKB and Fagerström Test for Nicotine Dependence from non-UKB cohorts.

RESULTS

In a large phenome-wide association study, we show that smoking consumption and dependence phenotypes show similar strongly negatively associations with a plethora of diseases, whereas alcohol consumption shows patterns of genetic association that diverge from those of alcohol misuse.

CONCLUSIONS

Our study suggests that cigarette smoking consumption, which can be easily measured in the general population, may be good a genetic proxy for nicotine dependence, whereas alcohol consumption is not a direct genetic proxy of alcohol misuse.

Polygenic risk score, healthy lifestyles, and risk of incident depression

Genetic factors increase the risk of depression, but the extent to which this can be offset by modifiable lifestyle factors is unknown. We investigated whether a combination of healthy lifestyles is associated with lower risk of depression regardless of genetic risk. Data were obtained from the UK Biobank and consisted of 339,767 participants (37-73 years old) without depression between 2006 and 2010. Genetic risk was categorized as low, intermediate, or high according to polygenic risk score for depression. A combination of healthy lifestyles factors-including no current smoking, regular physical activity, a healthy diet, moderate alcohol intake and a body mass index <30 kg/m²-was categorized into favorable, intermediate, and unfavorable lifestyles. The risk of depression was 22% higher among those at high genetic risk compared with those at low genetic risk (HR = 1.22, 95% CI: 1.14-1.30). Participants with high genetic risk and unfavorable lifestyle had a more than two-fold risk of incident depression compared with low genetic risk and favorable lifestyle (HR = 2.18, 95% CI: 1.84-2.58). There was no significant interaction between genetic risk and lifestyle factors (P for interaction = 0.69). Among participants at high genetic risk, a favorable lifestyle was associated with nearly 50% lower relative risk of depression than an unfavorable lifestyle (HR = 0.51, 95% CI: 0.43-0.60). We concluded that genetic and lifestyle factors were independently associated with risk of incident depression. Adherence to healthy lifestyles may lower the risk of depression regardless of genetic risk.

Elevated Polygenic Burden for Autism Spectrum Disorder Is Associated With the Broad Autism Phenotype in Mothers of Individuals With Autism Spectrum Disorder

BACKGROUND

Autism spectrum disorder (ASD) is a multifactorial neurodevelopmental disorder that encompasses a complex and heterogeneous set of traits. Subclinical traits that mirror the core features of ASD, referred to as the broad autism phenotype (BAP), have been documented repeatedly in unaffected relatives and are believed to reflect underlying genetic liability to ASD. The BAP may help inform the etiology of ASD by allowing the stratification of families into more phenotypically and etiologically homogeneous subgroups. This study explores polygenic scores related to the BAP.

METHODS

Phenotypic and genotypic information were obtained from 2614 trios from the Simons Simplex Collection. Polygenic scores of ASD (ASD-PGSs) were generated across the sample to determine the shared genetic overlap between the BAP and ASD. Maternal and paternal ASD-PGSs were explored in relation to BAP traits and their child's ASD symptomatology.

RESULTS

Maternal pragmatic language was related to child's social communicative atypicalities. In fathers, rigid personality was related to increased repetitive behaviors in children. Maternal (but not paternal) ASD-PGSs were related to the pragmatic language and rigid BAP domains.

CONCLUSIONS

Associations emerged between parent and child phenotypes, with more associations emerging in mothers than in fathers. ASD-PGS associations emerged with BAP in mothers only, highlighting the potential for a female protective factor, and implicating the polygenic etiology of ASD-related phenotypes in the BAP.

Polygenic Risk Scores Predict Hypertension Onset and Cardiovascular Risk

[Figure: see text].

Population-specific causal disease effect sizes in functionally important regions impacted by selection

Many diseases exhibit population-specific causal effect sizes with trans-ethnic genetic correlations significantly less than 1, limiting trans-ethnic polygenic risk prediction. We develop a new method, S-LDXR, for stratifying squared trans-ethnic genetic correlation across genomic annotations, and apply S-LDXR to genome-wide summary statistics for 31 diseases and complex traits in East Asians (average N = 90K) and Europeans (average N = 267K) with an average trans-ethnic genetic correlation of 0.85. We determine that squared trans-ethnic genetic correlation is 0.82× (s.e. 0.01) depleted in the top quintile of background selection statistic, implying more population-specific causal effect sizes. Accordingly, causal effect sizes are more population-specific in functionally important regions, including conserved and regulatory regions. In regions surrounding specifically expressed genes, causal effect sizes are most population-specific for skin and immune genes, and least population-specific for brain genes. Our results could potentially be explained by stronger gene-environment interaction at loci impacted by selection, particularly positive selection.

Challenges of Adjusting Single-Nucleotide Polymorphism Effect Sizes for Linkage Disequilibrium

BACKGROUND

Genome-wide association studies (GWAS) were successful in identifying SNPs showing association with disease, but their individual effect sizes are small and require large sample sizes to achieve statistical significance. Methods of post-GWAS analysis, including gene-based, gene-set and polygenic risk scores, combine the SNP effect sizes in an attempt to boost the power of the analyses. To avoid giving undue weight to SNPs in linkage disequilibrium (LD), the LD needs to be taken into account in these analyses.

OBJECTIVES

We review methods that attempt to adjust the effect sizes (β-coefficients) of summary statistics, instead of simple LD pruning.

METHODS

We subject LD adjustment approaches to a mathematical analysis, recognising Tikhonov regularisation as a framework for comparison.

RESULTS

Observing the similarity of the processes involved with the more straightforward Tikhonov-regularised ordinary least squares estimate for multivariate regression coefficients, we note that current methods based on a Bayesian model for the effect sizes effectively provide an implicit choice of the regularisation parameter, which is convenient, but at the price of reduced transparency and, especially in smaller LD blocks, a risk of incomplete LD correction.

CONCLUSIONS

There is no simple answer to the question which method is best, but where interpretability of the LD adjustment is essential, as in research aiming at identifying the genomic aetiology of disorders, our study suggests that a more direct choice of mild regularisation in the correction of effect sizes may be preferable.

Cigarette smoking behaviors and the importance of ethnicity and genetic ancestry

Cigarette smoking contributes to numerous diseases and is one of the leading causes of death in the United States. Smoking behaviors vary widely across race/ethnicity, but it is not clear why. Here, we examine the contribution of genetic ancestry to variation in two smoking-related traits in 43,485 individuals from four race/ethnicity groups (non-Hispanic white, Hispanic/Latino, East Asian, and African American) from a single U.S. healthcare plan. Smoking prevalence was the lowest among East Asians (22.7%) and the highest among non-Hispanic whites (38.5%). We observed significant associations between genetic ancestry and smoking-related traits. Within East Asians, we observed higher smoking prevalence with greater European (versus Asian) ancestry (P = 9.95 × 10^-12). Within Hispanic/Latinos, higher cigarettes per day (CPD) was associated with greater European ancestry (P = 3.34 × 10^-25). Within non-Hispanic whites, the lowest number of CPD was observed for individuals of southeastern European ancestry (P = 9.06 × 10^-5). These associations remained after considering known smoking-associated loci, education, socioeconomic factors, and marital status. Our findings support the role of genetic ancestry and socioeconomic factors in cigarette smoking behaviors in non-Hispanic whites, Hispanic/Latinos, and East Asians.

Polygenic risk modeling with latent trait-related genetic components

Polygenic risk models have led to significant advances in understanding complex diseases and their clinical presentation. While polygenic risk scores (PRS) can effectively predict outcomes, they do not generally account for disease subtypes or pathways which underlie within-trait diversity. Here, we introduce a latent factor model of genetic risk based on components from Decomposition of Genetic Associations (DeGAs), which we call the DeGAs polygenic risk score (dPRS). We compute DeGAs using genetic associations for 977 traits and find that dPRS performs comparably to standard PRS while offering greater interpretability. We show how to decompose an individual's genetic risk for a trait across DeGAs components, with examples for body mass index (BMI) and myocardial infarction (heart attack) in 337,151 white British individuals in the UK Biobank, with replication in a further set of 25,486 non-British white individuals. We find that BMI polygenic risk factorizes into components related to fat-free mass, fat mass, and overall health indicators like physical activity. Most individuals with high dPRS for BMI have strong contributions from both a fat-mass component and a fat-free mass component, whereas a few "outlier" individuals have strong contributions from only one of the two components. Overall, our method enables fine-scale interpretation of the drivers of genetic risk for complex traits.

Clinical laboratory test-wide association scan of polygenic scores identifies biomarkers of complex disease

BACKGROUND

Clinical laboratory (lab) tests are used in clinical practice to diagnose, treat, and monitor disease conditions. Test results are stored in electronic health records (EHRs), and a growing number of EHRs are linked to patient DNA, offering unprecedented opportunities to query relationships between genetic risk for complex disease and quantitative physiological measurements collected on large populations.

METHODS

A total of 3075 quantitative lab tests were extracted from Vanderbilt University Medical Center's (VUMC) EHR system and cleaned for population-level analysis according to our QualityLab protocol. Lab values extracted from BioVU were compared with previous population studies using heritability and genetic correlation analyses. We then tested the hypothesis that polygenic risk scores for biomarkers and complex disease are associated with biomarkers of disease extracted from the EHR. In a proof of concept analyses, we focused on lipids and coronary artery disease (CAD). We cleaned lab traits extracted from the EHR performed lab-wide association scans (LabWAS) of the lipids and CAD polygenic risk scores across 315 heritable lab tests then replicated the pipeline and analyses in the Massachusetts General Brigham Biobank.

RESULTS

Heritability estimates of lipid values (after cleaning with QualityLab) were comparable to previous reports and polygenic scores for lipids were strongly associated with their referent lipid in a LabWAS. LabWAS of the polygenic score for CAD recapitulated canonical heart disease biomarker profiles including decreased HDL, increased pre-medication LDL, triglycerides, blood glucose, and glycated hemoglobin (HgbA1C) in European and African descent populations. Notably, many of these associations remained even after adjusting for the presence of cardiovascular disease and were replicated in the MGBB.

CONCLUSIONS

Polygenic risk scores can be used to identify biomarkers of complex disease in large-scale EHR-based genomic analyses, providing new avenues for discovery of novel biomarkers and deeper understanding of disease trajectories in pre-symptomatic individuals. We present two methods and associated software, QualityLab and LabWAS, to clean and analyze EHR labs at scale and perform a Lab-Wide Association Scan.

Interactions Between Enhanced Polygenic Risk Scores and Lifestyle for Cardiovascular Disease, Diabetes, and Lipid Levels

BACKGROUND

Both lifestyle and genetic factors confer risk for cardiovascular diseases, type 2 diabetes, and dyslipidemia. However, the interactions between these 2 groups of risk factors were not comprehensively understood due to previous poor estimation of genetic risk. Here we set out to develop enhanced polygenic risk scores (PRS) and systematically investigate multiplicative and additive interactions between PRS and lifestyle for coronary artery disease, atrial fibrillation, type 2 diabetes, total cholesterol, triglyceride, and LDL-cholesterol.

METHODS

Our study included 276 096 unrelated White British participants from the UK Biobank. We investigated several PRS methods (P+T, LDpred, PRS continuous shrinkage, and AnnoPred) and showed that AnnoPred achieved consistently improved prediction accuracy for all 6 diseases/traits. With enhanced PRS and combined lifestyle status categorized by smoking, body mass index, physical activity, and diet, we investigated both multiplicative and additive interactions between PRS and lifestyle using regression models.

RESULTS

We observed that healthy lifestyle reduced disease incidence by similar multiplicative magnitude across different PRS groups. The absolute risk reduction from lifestyle adherence was, however, significantly greater in individuals with higher PRS. Specifically, for type 2 diabetes, the absolute risk reduction from lifestyle adherence was 12.4% (95% CI, 10.0%-14.9%) in the top 1% PRS versus 2.8% (95% CI, 2.3%-3.3%) in the bottom PRS decile, leading to a ratio of >4.4. We also observed a significant interaction effect between PRS and lifestyle on triglyceride level.

CONCLUSIONS

By leveraging functional annotations, AnnoPred outperforms state-of-the-art methods on quantifying genetic risk through PRS. Our analyses based on enhanced PRS suggest that individuals with high genetic risk may derive similar relative but greater absolute benefit from lifestyle adherence.

Clinical and genetic differences between bipolar disorder type 1 and 2 in multiplex families

The two major subtypes of bipolar disorder (BD), BD-I and BD-II, are distinguished based on the presence of manic or hypomanic episodes. Historically, BD-II was perceived as a less severe form of BD-I. Recent research has challenged this concept of a severity continuum. Studies in large samples of unrelated patients have described clinical and genetic differences between the subtypes. Besides an increased schizophrenia polygenic risk load in BD-I, these studies also observed an increased depression risk load in BD-II patients. The present study assessed whether such clinical and genetic differences are also found in BD patients from multiplex families, which exhibit reduced genetic and environmental heterogeneity. Comparing 252 BD-I and 75 BD-II patients from the Andalusian Bipolar Family (ABiF) study, the clinical course, symptoms during depressive and manic episodes, and psychiatric comorbidities were analyzed. Furthermore, polygenic risk scores (PRS) for BD, schizophrenia, and depression were assessed. BD-I patients not only suffered from more severe symptoms during manic episodes but also more frequently showed incapacity during depressive episodes. A higher BD PRS was significantly associated with suicidal ideation. Moreover, BD-I cases exhibited lower depression PRS. In line with a severity continuum from BD-II to BD-I, our results link BD-I to a more pronounced clinical presentation in both mania and depression and indicate that the polygenic risk load of BD predisposes to more severe disorder characteristics. Nevertheless, our results suggest that the genetic risk burden for depression also shapes disorder presentation and increases the likelihood of BD-II subtype development.

Genome-wide association study of problematic opioid prescription use in 132,113 23andMe research participants of European ancestry

The growing prevalence of opioid use disorder (OUD) constitutes an urgent health crisis. Ample evidence indicates that risk for OUD is heritable. As a surrogate (or proxy) for OUD, we explored the genetic basis of using prescription opioids 'not as prescribed'. We hypothesized that misuse of opiates might be a heritable risk factor for OUD. To test this hypothesis, we performed a genome-wide association study (GWAS) of problematic opioid use (POU) in 23andMe research participants of European ancestry (N = 132,113; 21% cases). We identified two genome-wide significant loci (rs3791033, an intronic variant of KDM4A; rs640561, an intergenic variant near LRRIQ3). POU showed positive genetic correlations with the two largest available GWAS of OUD and opioid dependence (r_g = 0.64, 0.80, respectively). We also identified numerous additional genetic correlations with POU, including alcohol dependence (r_g = 0.74), smoking initiation (r_g = 0.63), pain relief medication intake (r_g = 0.49), major depressive disorder (r_g = 0.44), chronic pain (r_g = 0.42), insomnia (r_g = 0.39), and loneliness (r_g = 0.28). Although POU was positively genetically correlated with risk-taking (r_g = 0.38), conditioning POU on risk-taking did not substantially alter the magnitude or direction of these genetic correlations, suggesting that POU does not simply reflect a genetic tendency towards risky behavior. Lastly, we performed phenome- and lab-wide association analyses, which uncovered additional phenotypes that were associated with POU, including respiratory failure, insomnia, ischemic heart disease, and metabolic and blood-related biomarkers. We conclude that opioid misuse can be measured in population-based cohorts and provides a cost-effective complementary strategy for understanding the genetic basis of OUD.

Characterizing the Clinical and Genetic Spectrum of Polycystic Ovary Syndrome in Electronic Health Records

CONTEXT

Polycystic ovary syndrome (PCOS) is one of the leading causes of infertility, yet current diagnostic criteria are ineffective at identifying patients whose symptoms reside outside strict diagnostic criteria. As a result, PCOS is underdiagnosed and its etiology is poorly understood.

OBJECTIVE

We aim to characterize the phenotypic spectrum of PCOS clinical features within and across racial and ethnic groups.

METHODS

We developed a strictly defined PCOS algorithm (PCOSkeyword-strict) using the International Classification of Diseases, ninth and tenth revisions and keywords mined from clinical notes in electronic health records (EHRs) data. We then systematically relaxed the inclusion criteria to evaluate the change in epidemiological and genetic associations resulting in 3 subsequent algorithms (PCOScoded-broad, PCOScoded-strict, and PCOSkeyword-broad). We evaluated the performance of each phenotyping approach and characterized prominent clinical features observed in racially and ethnically diverse PCOS patients.

RESULTS

The best performance came from the PCOScoded-strict algorithm, with a positive predictive value of 98%. Individuals classified as cases by this algorithm had significantly higher body mass index (BMI), insulin levels, free testosterone values, and genetic risk scores for PCOS, compared to controls. Median BMI was higher in African American females with PCOS compared to White and Hispanic females with PCOS.

CONCLUSIONS

PCOS symptoms are observed across a severity spectrum that parallels the continuous genetic liability to PCOS in the general population. Racial and ethnic group differences exist in PCOS symptomology and metabolic health across different phenotyping strategies.

LDpred2: better, faster, stronger

Motivation

Polygenic scores have become a central tool in human genetics research. LDpred is a popular method for deriving polygenic scores based on summary statistics and a matrix of correlation between genetic variants. However, LDpred has limitations that may reduce its predictive performance.

Results

Here, we present LDpred2, a new version of LDpred that addresses these issues. We also provide two new options in LDpred2: a ‘sparse’ option that can learn effects that are exactly 0, and an ‘auto’ option that directly learns the two LDpred parameters from data. We benchmark predictive performance of LDpred2 against the previous version on simulated and real data, demonstrating substantial improvements in robustness and predictive accuracy compared to LDpred1. We then show that LDpred2 also outperforms other polygenic score methods recently developed, with a mean AUC over the 8 real traits analyzed here of 65.1%, compared to 63.8% for lassosum, 62.9% for PRS-CS and 61.5% for SBayesR. Note that LDpred2 provides more accurate polygenic scores when run genome-wide, instead of per chromosome.

Availability and implementation

LDpred2 is implemented in R package bigsnpr.

Supplementary information

Supplementary data are available at Bioinformatics online.

The role of polygenic risk and susceptibility genes in breast cancer over the course of life

Polygenic risk scores (PRS) for breast cancer have potential to improve risk prediction, but there is limited information on their utility in various clinical situations. Here we show that among 122,978 women in the FinnGen study with 8401 breast cancer cases, the PRS modifies the breast cancer risk of two high-impact frameshift risk variants. Similarly, we show that after the breast cancer diagnosis, individuals with elevated PRS have an elevated risk of developing contralateral breast cancer, and that the PRS can considerably improve risk assessment among their female first-degree relatives. In more detail, women with the c.1592delT variant in PALB2 (242-fold enrichment in Finland, 336 carriers) and an average PRS (10–90^th percentile) have a lifetime risk of breast cancer at 55% (95% CI 49–61%), which increases to 84% (71–97%) with a high PRS ( > 90^th percentile), and decreases to 49% (30–68%) with a low PRS ( < 10^th percentile). Similarly, for c.1100delC in CHEK2 (3.7–fold enrichment; 1648 carriers), the respective lifetime risks are 29% (27–32%), 59% (52–66%), and 9% (5–14%). The PRS also refines the risk assessment of women with first-degree relatives diagnosed with breast cancer, particularly among women with positive family history of early-onset breast cancer. Here we demonstrate the opportunities for a comprehensive way of assessing genetic risk in the general population, in breast cancer patients, and in unaffected family members.

Identifying women at high risk of breast cancer has important implications for screening. Here, the authors demonstrate that polygenic risk scores improve breast cancer risk prediction in the population, in women with mutations in high-risk genes and in women with close relatives with the disease.

A large-scale genome-wide association study meta-analysis of cannabis use disorder

BACKGROUND

Variation in liability to cannabis use disorder has a strong genetic component (estimated twin and family heritability about 50-70%) and is associated with negative outcomes, including increased risk of psychopathology. The aim of the study was to conduct a large genome-wide association study (GWAS) to identify novel genetic variants associated with cannabis use disorder.

METHODS

To conduct this GWAS meta-analysis of cannabis use disorder and identify associations with genetic loci, we used samples from the Psychiatric Genomics Consortium Substance Use Disorders working group, iPSYCH, and deCODE (20 916 case samples, 363 116 control samples in total), contrasting cannabis use disorder cases with controls. To examine the genetic overlap between cannabis use disorder and 22 traits of interest (chosen because of previously published phenotypic correlations [eg, psychiatric disorders] or hypothesised associations [eg, chronotype] with cannabis use disorder), we used linkage disequilibrium score regression to calculate genetic correlations.

FINDINGS

We identified two genome-wide significant loci: a novel chromosome 7 locus (FOXP2, lead single-nucleotide polymorphism [SNP] rs7783012; odds ratio [OR] 1·11, 95% CI 1·07-1·15, p=1·84 × 10-9) and the previously identified chromosome 8 locus (near CHRNA2 and EPHX2, lead SNP rs4732724; OR 0·89, 95% CI 0·86-0·93, p=6·46 × 10-9). Cannabis use disorder and cannabis use were genetically correlated (rg 0·50, p=1·50 × 10-21), but they showed significantly different genetic correlations with 12 of the 22 traits we tested, suggesting at least partially different genetic underpinnings of cannabis use and cannabis use disorder. Cannabis use disorder was positively genetically correlated with other psychopathology, including ADHD, major depression, and schizophrenia.

INTERPRETATION

These findings support the theory that cannabis use disorder has shared genetic liability with other psychopathology, and there is a distinction between genetic liability to cannabis use and cannabis use disorder.

FUNDING

National Institute of Mental Health; National Institute on Alcohol Abuse and Alcoholism; National Institute on Drug Abuse; Center for Genomics and Personalized Medicine and the Centre for Integrative Sequencing; The European Commission, Horizon 2020; National Institute of Child Health and Human Development; Health Research Council of New Zealand; National Institute on Aging; Wellcome Trust Case Control Consortium; UK Research and Innovation Medical Research Council (UKRI MRC); The Brain & Behavior Research Foundation; National Institute on Deafness and Other Communication Disorders; Substance Abuse and Mental Health Services Administration (SAMHSA); National Institute of Biomedical Imaging and Bioengineering; National Health and Medical Research Council (NHMRC) Australia; Tobacco-Related Disease Research Program of the University of California; Families for Borderline Personality Disorder Research (Beth and Rob Elliott) 2018 NARSAD Young Investigator Grant; The National Child Health Research Foundation (Cure Kids); The Canterbury Medical Research Foundation; The New Zealand Lottery Grants Board; The University of Otago; The Carney Centre for Pharmacogenomics; The James Hume Bequest Fund; National Institutes of Health: Genes, Environment and Health Initiative; National Institutes of Health; National Cancer Institute; The William T Grant Foundation; Australian Research Council; The Virginia Tobacco Settlement Foundation; The VISN 1 and VISN 4 Mental Illness Research, Education, and Clinical Centers of the US Department of Veterans Affairs; The 5th Framework Programme (FP-5) GenomEUtwin Project; The Lundbeck Foundation; NIH-funded Shared Instrumentation Grant S10RR025141; Clinical Translational Sciences Award grants; National Institute of Neurological Disorders and Stroke; National Heart, Lung, and Blood Institute; National Institute of General Medical Sciences.

Cancer PRSweb: An Online Repository with Polygenic Risk Scores for Major Cancer Traits and Their Evaluation in Two Independent Biobanks

To facilitate scientific collaboration on polygenic risk scores (PRSs) research, we created an extensive PRS online repository for 35 common cancer traits integrating freely available genome-wide association studies (GWASs) summary statistics from three sources: published GWASs, the NHGRI-EBI GWAS Catalog, and UK Biobank-based GWASs. Our framework condenses these summary statistics into PRSs using various approaches such as linkage disequilibrium pruning/p value thresholding (fixed or data-adaptively optimized thresholds) and penalized, genome-wide effect size weighting. We evaluated the PRSs in two biobanks: the Michigan Genomics Initiative (MGI), a longitudinal biorepository effort at Michigan Medicine, and the population-based UK Biobank (UKB). For each PRS construct, we provide measures on predictive performance and discrimination. Besides PRS evaluation, the Cancer-PRSweb platform features construct downloads and phenome-wide PRS association study results (PRS-PheWAS) for predictive PRSs. We expect this integrated platform to accelerate PRS-related cancer research.

Risk Stratification for Bipolar Disorder Using Polygenic Risk Scores Among Young High-Risk Adults

Objective: Identifying high-risk groups with an increased genetic liability for bipolar disorder (BD) will provide insights into the etiology of BD and contribute to early detection of BD. We used the BD polygenic risk score (PRS) derived from BD genome-wide association studies (GWAS) to explore how such genetic risk manifests in young, high-risk adults. We postulated that BD-PRS would be associated with risk factors for BD. Methods: A final sample of 185 young, high-risk German adults (aged 18-35 years) were grouped into three risk groups and compared to a healthy control group (n = 1,100). The risk groups comprised 117 cases with attention deficit hyperactivity disorder (ADHD), 45 with major depressive disorder (MDD), and 23 help-seeking adults with early recognition symptoms [ER: positive family history for BD, (sub)threshold affective symptomatology and/or mood swings, sleeping disorder]. BD-PRS was computed for each participant. Logistic regression models (controlling for sex, age, and the first five ancestry principal components) were used to assess associations of BD-PRS and the high-risk phenotypes. Results: We observed an association between BD-PRS and combined risk group status (OR = 1.48, p < 0.001), ADHD diagnosis (OR = 1.32, p = 0.009), MDD diagnosis (OR = 1.96, p < 0.001), and ER group status (OR = 1.7, p = 0.025; not significant after correction for multiple testing) compared to healthy controls. Conclusions: In the present study, increased genetic risk for BD was a significant predictor for MDD and ADHD status, but not for ER. These findings support an underlying shared risk for both MDD and BD as well as ADHD and BD. Improving our understanding of the underlying genetic architecture of these phenotypes may aid in early identification and risk stratification.

A fast and scalable framework for large-scale and ultrahigh-dimensional sparse regression with application to the UK Biobank

The UK Biobank is a very large, prospective population-based cohort study across the United Kingdom. It provides unprecedented opportunities for researchers to investigate the relationship between genotypic information and phenotypes of interest. Multiple regression methods, compared with genome-wide association studies (GWAS), have already been showed to greatly improve the prediction performance for a variety of phenotypes. In the high-dimensional settings, the lasso, since its first proposal in statistics, has been proved to be an effective method for simultaneous variable selection and estimation. However, the large-scale and ultrahigh dimension seen in the UK Biobank pose new challenges for applying the lasso method, as many existing algorithms and their implementations are not scalable to large applications. In this paper, we propose a computational framework called batch screening iterative lasso (BASIL) that can take advantage of any existing lasso solver and easily build a scalable solution for very large data, including those that are larger than the memory size. We introduce snpnet, an R package that implements the proposed algorithm on top of glmnet and optimizes for single nucleotide polymorphism (SNP) datasets. It currently supports ℓ1-penalized linear model, logistic regression, Cox model, and also extends to the elastic net with ℓ1/ℓ2 penalty. We demonstrate results on the UK Biobank dataset, where we achieve competitive predictive performance for all four phenotypes considered (height, body mass index, asthma, high cholesterol) using only a small fraction of the variants compared with other established polygenic risk score methods.

Deep neural network improves the estimation of polygenic risk scores for breast cancer

Polygenic risk scores (PRS) estimate the genetic risk of an individual for a complex disease based on many genetic variants across the whole genome. In this study, we compared a series of computational models for estimation of breast cancer PRS. A deep neural network (DNN) was found to outperform alternative machine learning techniques and established statistical algorithms, including BLUP, BayesA, and LDpred. In the test cohort with 50% prevalence, the Area Under the receiver operating characteristic Curve (AUC) were 67.4% for DNN, 64.2% for BLUP, 64.5% for BayesA, and 62.4% for LDpred. BLUP, BayesA, and LPpred all generated PRS that followed a normal distribution in the case population. However, the PRS generated by DNN in the case population followed a bimodal distribution composed of two normal distributions with distinctly different means. This suggests that DNN was able to separate the case population into a high-genetic-risk case subpopulation with an average PRS significantly higher than the control population and a normal-genetic-risk case subpopulation with an average PRS similar to the control population. This allowed DNN to achieve 18.8% recall at 90% precision in the test cohort with 50% prevalence, which can be extrapolated to 65.4% recall at 20% precision in a general population with 12% prevalence. Interpretation of the DNN model identified salient variants that were assigned insignificant p values by association studies, but were important for DNN prediction. These variants may be associated with the phenotype through nonlinear relationships.