Learning Gaussian Graphical Models from Correlated Data.

Gaussian Graphical Models (GGM) have been widely used in biomedical research to explore complex relationships between many variables. There are well established procedures to build GGMs from a sample of independent and identical distributed observations. However, many studies include clustered and longitudinal data that result in correlated observations and ignoring this correlation among observations can lead to inflated Type I error. In this paper, we propose a Bootstrap algorithm to infer GGM from correlated data. We use extensive simulations of correlated data from family-based studies to show that the Bootstrap method does not inflate the Type I error while retaining statistical power compared to alternative solutions. We apply our method to learn the GGM that represents complex relations between 47 Polygenic Risk Scores generated using genome-wide genotype data from a family-based study known as the Long Life Family Study. By comparing it to the conventional methods that ignore within-cluster correlation, we show that our method controls the Type I error well in this real example.

Correlation-based tests for the formal comparison of polygenic scores in multiple populations.

Polygenic scores (PGS) are measures of genetic risk, derived from the results of genome wide association studies (GWAS). Previous work has proposed the coefficient of determination (R2) as an appropriate measure by which to compare PGS performance in a validation dataset. Here we propose correlation-based methods for evaluating PGS performance by adapting previous work which produced a statistical framework and robust test statistics for the comparison of multiple correlation measures in multiple populations. This flexible framework can be extended to a wider variety of hypothesis tests than currently available methods. We assess our proposed method in simulation and demonstrate its utility with two examples, assessing previously developed PGS for low-density lipoprotein cholesterol and height in multiple populations in the All of Us cohort. Finally, we provide an R package 'coranova' with both parametric and nonparametric implementations of the described methods.

The relationship between 11 different polygenic longevity scores, parental lifespan, and disease diagnosis in the UK Biobank.

Large-scale genome-wide association studies (GWAS) strongly suggest that most traits and diseases have a polygenic component. This observation has motivated the development of disease-specific "polygenic scores (PGS)" that are weighted sums of the effects of disease-associated variants identified from GWAS that correlate with an individual's likelihood of expressing a specific phenotype. Although most GWAS have been pursued on disease traits, leading to the creation of refined "Polygenic Risk Scores" (PRS) that quantify risk to diseases, many GWAS have also been pursued on extreme human longevity, general fitness, health span, and other health-positive traits. These GWAS have discovered many genetic variants seemingly protective from disease and are often different from disease-associated variants (i.e., they are not just alternative alleles at disease-associated loci) and suggest that many health-positive traits also have a polygenic basis. This observation has led to an interest in "polygenic longevity scores (PLS)" that quantify the "risk" or genetic predisposition of an individual towards health. We derived 11 different PLS from 4 different available GWAS on lifespan and then investigated the properties of these PLS using data from the UK Biobank (UKB). Tests of association between the PLS and population structure, parental lifespan, and several cancerous and non-cancerous diseases, including death from COVID-19, were performed. Based on the results of our analyses, we argue that PLS are made up of variants not only robustly associated with parental lifespan, but that also contribute to the genetic architecture of disease susceptibility, morbidity, and mortality.

Clinical and Genetic Atrial Fibrillation Risk and Discrimination of Cardioembolic From Noncardioembolic Stroke.

BACKGROUND: Stroke is a leading cause of death and disability worldwide. Atrial fibrillation (AF) is a common cause of stroke but may not be detectable at the time of stroke. We hypothesized that an AF polygenic risk score (PRS) can discriminate between cardioembolic stroke and noncardioembolic strokes. METHODS: We evaluated AF and stroke risk in 26 145 individuals of European descent from the Stroke Genetics Network case-control study. AF genetic risk was estimated using 3 recently developed PRS methods (LDpred-funct-inf, sBayesR, and PRS-CS) and 2 previously validated PRSs. We performed logistic regression of each AF PRS on AF status and separately cardioembolic stroke, adjusting for clinical risk score (CRS), imputation group, and principal components. We calculated model discrimination of AF and cardioembolic stroke using the concordance statistic (c-statistic) and compared c-statistics using 2000-iteration bootstrapping. We also assessed reclassification of cardioembolic stroke with the addition of PRS to either CRS or a modified CHA2DS2-VASc score alone. RESULTS: Each AF PRS was significantly associated with AF and with cardioembolic stroke after adjustment for CRS. Addition of each AF PRS significantly improved discrimination as compared with CRS alone (P<0.01). When combined with the CRS, both PRS-CS and LDpred scores discriminated both AF and cardioembolic stroke (c-statistic 0.84 for AF; 0.74 for cardioembolic stroke) better than 3 other PRS scores (P<0.01). Using PRS-CS PRS and CRS in combination resulted in more appropriate reclassification of stroke events as compared with CRS alone (event reclassification [net reclassification indices]+=14% [95% CI, 10%-18%]; nonevent reclassification [net reclassification indices]-=17% [95% CI, 15%-0.19%]) or the modified CHA2DS2-VASc score (net reclassification indices+=11% [95% CI, 7%-15%]; net reclassification indices-=14% [95% CI, 12%-16%]) alone. CONCLUSIONS: Addition of polygenic risk of AF to clinical risk factors modestly improves the discrimination of cardioembolic from noncardioembolic strokes, as well as reclassification of stroke subtype. Polygenic risk of AF may be a useful biomarker for identifying strokes caused by AF.

Signatures of Neuropsychological Test Results in the Long Life Family Study: A Cluster Analysis.

BACKGROUND: Discovering patterns of cognitive domains and characterizing how these patterns associate with other risk factors and biomarkers can improve our understanding of the determinants of cognitive aging. OBJECTIVE: To discover patterns of cognitive domains using neuropsychological test results in Long Life Family Study (LLFS) and characterize how these patterns associate with aging markers. METHODS: 5,086 LLFS participants were administered neuropsychological tests at enrollment. We performed a cluster analysis of six baseline neuropsychological test scores and tested the association between the identified clusters and various clinical variables, biomarkers, and polygenic risk scores using generalized estimating equations and the Chi-square test. We used Cox regression to correlate the clusters with the hazard of various medical events. We investigated whether the cluster information could enhance the prediction of cognitive decline using Bayesian beta regression. RESULTS: We identified 12 clusters with different cognitive signatures that represent profiles of performance across multiple neuropsychological tests. These signatures significantly correlated with 26 variables including polygenic risk scores, physical and pulmonary functions, and blood biomarkers and were associated with the hazard of mortality (p < 0.01), cardiovascular disease (p = 0.03), dementia (p = 0.01), and skin cancer (p = 0.03). CONCLUSION: The identified cognitive signatures capture multiple domains simultaneously and provide a holistic vision of cognitive function, showing that different patterns of cognitive function can coexist in aging individuals. Such patterns can be used for clinical intervention and primary care.

Endophenotype effect sizes support variant pathogenicity in monogenic disease susceptibility genes.

Accurate and efficient classification of variant pathogenicity is critical for research and clinical care. Using data from three large studies, we demonstrate that population-based associations between rare variants and quantitative endophenotypes for three monogenic diseases (low-density-lipoprotein cholesterol for familial hypercholesterolemia, electrocardiographic QTc interval for long QT syndrome, and glycosylated hemoglobin for maturity-onset diabetes of the young) provide evidence for variant pathogenicity. Effect sizes are associated with pathogenic ClinVar assertions (P < 0.001 for each trait) and discriminate pathogenic from non-pathogenic variants (area under the curve 0.82-0.84 across endophenotypes). An effect size threshold of ≥ 0.5 times the endophenotype standard deviation nominates up to 35% of rare variants of uncertain significance or not in ClinVar in disease susceptibility genes with pathogenic potential. We propose that variant associations with quantitative endophenotypes for monogenic diseases can provide evidence supporting pathogenicity.

Defining Sepsis Phenotypes-Two Murine Models of Sepsis and Machine Learning.

INTRODUCTION: The immunobiology defining the clinically apparent differences in response to sepsis remains unclear. We hypothesize that in murine models of sepsis we can identify phenotypes of sepsis using non-invasive physiologic parameters (NIPP) early after infection to distinguish between different inflammatory states. METHODS: Two murine models of sepsis were used: gram-negative pneumonia (PNA) and cecal ligation and puncture (CLP). All mice were treated with broad spectrum antibiotics and fluid resuscitation. High-risk sepsis responders (pDie) were defined as those predicted to die within 72 h following infection. Low-risk responders (pLive) were expected to survive the initial 72 h of sepsis. Statistical modeling in R was used for statistical analysis and machine learning. RESULTS: NIPP obtained at 6 and 24 h after infection of 291 mice (85 PNA and 206 CLP) were used to define the sepsis phenotypes. Lasso regression for variable selection with 10-fold cross-validation was used to define the optimal shrinkage parameters. The variables selected to discriminate between phenotypes included 6-h temperature and 24-h pulse distention, heart rate (HR), and temperature. Applying the model to fit test data (n = 55), area under the curve (AUC) for the receiver operating characteristics (ROC) curve was 0.93. Subgroup analysis of 120 CLP mice revealed a HR of <620 bpm at 24 h as a univariate predictor of pDie. (AUC of ROC curve = 0.90). Subgroup analysis of PNA exposed mice (n = 121) did not reveal a single predictive variable highlighting the complex physiological alterations in response to sepsis. CONCLUSION: In murine models with various etiologies of sepsis, non-invasive vitals assessed just 6 and 24 h after infection can identify different sepsis phenotypes. Stratification by sepsis phenotypes can transform future studies investigating novel therapies for sepsis.

Distribution of 54 polygenic risk scores for common diseases in long lived individuals and their offspring.

A surprising and well-replicated result in genetic studies of human longevity is that centenarians appear to carry disease-associated variants in numbers similar to the general population. With the proliferation of large genome-wide association studies (GWAS) in recent years, investigators have turned to polygenic scores to leverage GWAS results into a measure of genetic risk that can better predict the risk of disease than individual significant variants alone. We selected 54 polygenic risk scores (PRSs) developed for a variety of outcomes, and we calculated their values in individuals from the New England Centenarian Study (NECS, N = 4886) and the Long Life Family Study (LLFS, N = 4577). We compared the distribution of these PRSs among exceptionally long-lived individuals (ELLI), their offspring, and controls, and we also examined their predictive values, using t-tests and regression models adjusting for sex and principal components reflecting the ancestral background of the individuals (PCs). In our analyses, we controlled for multiple testing using a Bonferroni-adjusted threshold for 54 traits. We found that only 4 of the 54 PRSs differed between ELLIs and controls in both cohorts. ELLIs had significantly lower mean PRSs for Alzheimer's disease (AD) and coronary artery disease (CAD) than controls, suggesting a genetic predisposition to extreme longevity may be mediated by reduced susceptibility to these traits. ELLIs also had significantly higher mean PRSs for improved cognitive function and parental extreme longevity. In addition, the PRS for AD was associated with a higher risk of dementia among controls but not ELLIs (p = 0.003, 0.3 in NECS, p = 0.03, 0.9 in LLFS, respectively). ELLIs have a similar burden of genetic disease risk as the general population for most traits but have a significantly lower genetic risk of AD and CAD. The lack of association between AD PRS and dementia among ELLIs suggests that the genetic risk for AD that they do have is somehow counteracted by protective genetic or environmental factors.