Adjusting for genetic confounders in transcriptome-wide association studies improves discovery of risk genes of complex traits.

Many methods have been developed to leverage expression quantitative trait loci (eQTL) data to nominate candidate genes from genome-wide association studies. These methods, including colocalization, transcriptome-wide association studies (TWAS) and Mendelian randomization-based methods; however, all suffer from a key problem-when assessing the role of a gene in a trait using its eQTLs, nearby variants and genetic components of other genes' expression may be correlated with these eQTLs and have direct effects on the trait, acting as potential confounders. Our extensive simulations showed that existing methods fail to account for these 'genetic confounders', resulting in severe inflation of false positives. Our new method, causal-TWAS (cTWAS), borrows ideas from statistical fine-mapping and allows us to adjust all genetic confounders. cTWAS showed calibrated false discovery rates in simulations, and its application on several common traits discovered new candidate genes. In conclusion, cTWAS provides a robust statistical framework for gene discovery.

Inferring cell-type-specific causal gene regulatory networks during human neurogenesis.

BACKGROUND: Genetic variation influences both chromatin accessibility, assessed in chromatin accessibility quantitative trait loci (caQTL) studies, and gene expression, assessed in expression QTL (eQTL) studies. Genetic variants can impact either nearby genes (cis-eQTLs) or distal genes (trans-eQTLs). Colocalization between caQTL and eQTL, or cis- and trans-eQTLs suggests that they share causal variants. However, pairwise colocalization between these molecular QTLs does not guarantee a causal relationship. Mediation analysis can be applied to assess the evidence supporting causality versus independence between molecular QTLs. Given that the function of QTLs can be cell-type-specific, we performed mediation analyses to find epigenetic and distal regulatory causal pathways for genes within two major cell types of the developing human cortex, progenitors and neurons. RESULTS: We find that the expression of 168 and 38 genes is mediated by chromatin accessibility in progenitors and neurons, respectively. We also find that the expression of 11 and 12 downstream genes is mediated by upstream genes in progenitors and neurons. Moreover, we discover that a genetic locus associated with inter-individual differences in brain structure shows evidence for mediation of SLC26A7 through chromatin accessibility, identifying molecular mechanisms of a common variant association to a brain trait. CONCLUSIONS: In this study, we identify cell-type-specific causal gene regulatory networks whereby the impacts of variants on gene expression were mediated by chromatin accessibility or distal gene expression. Identification of these causal paths will enable identifying and prioritizing actionable regulatory targets perturbing these key processes during neurodevelopment.

Genetic Mapping of Multiple Traits Identifies Novel Genes for Adiposity, Lipids, and Insulin Secretory Capacity in Outbred Rats.

Despite the successes of human genome-wide association studies, the causal genes underlying most metabolic traits remain unclear. We used outbred heterogeneous stock (HS) rats, coupled with expression data and mediation analysis, to identify quantitative trait loci (QTLs) and candidate gene mediators for adiposity, glucose tolerance, serum lipids, and other metabolic traits. Physiological traits were measured in 1,519 male HS rats, with liver and adipose transcriptomes measured in >410 rats. Genotypes were imputed from low-coverage whole-genome sequencing. Linear mixed models were used to detect physiological and expression QTLs (pQTLs and eQTLs, respectively), using both single nucleotide polymorphism (SNP)- and haplotype-based models for pQTL mapping. Genes with cis-eQTLs that overlapped pQTLs were assessed as causal candidates through mediation analysis. We identified 14 SNP-based pQTLs and 19 haplotype-based pQTLs, of which 10 were in common. Using mediation, we identified the following genes as candidate mediators of pQTLs: Grk5 for fat pad weight and serum triglyceride pQTLs on Chr1, Krtcap3 for fat pad weight and serum triglyceride pQTLs on Chr6, Ilrun for a fat pad weight pQTL on Chr20, and Rfx6 for a whole pancreatic insulin content pQTL on Chr20. Furthermore, we verified Grk5 and Ktrcap3 using gene knockdown/out models, thereby shedding light on novel regulators of obesity.

A Bayesian model selection approach to mediation analysis.

Genetic studies often seek to establish a causal chain of events originating from genetic variation through to molecular and clinical phenotypes. When multiple phenotypes share a common genetic association, one phenotype may act as an intermediate for the genetic effects on the other. Alternatively, the phenotypes may be causally unrelated but share genetic loci. Mediation analysis represents a class of causal inference approaches used to determine which of these scenarios is most plausible. We have developed a general approach to mediation analysis based on Bayesian model selection and have implemented it in an R package, bmediatR. Bayesian model selection provides a flexible framework that can be tailored to different analyses. Our approach can incorporate prior information about the likelihood of models and the strength of causal effects. It can also accommodate multiple genetic variants or multi-state haplotypes. Our approach reports posterior probabilities that can be useful in interpreting uncertainty among competing models. We compared bmediatR with other popular methods, including the Sobel test, Mendelian randomization, and Bayesian network analysis using simulated data. We found that bmediatR performed as well or better than these alternatives in most scenarios. We applied bmediatR to proteome data from Diversity Outbred (DO) mice, a multi-parent population, and demonstrate the power of mediation with multi-state haplotypes. We also applied bmediatR to data from human cell lines to identify transcripts that are mediated through or are expressed independently from local chromatin accessibility. We demonstrate that Bayesian model selection provides a powerful and versatile approach to identify causal relationships in genetic studies using model organism or human data.

Transcriptome-wide analyses of adipose tissue in outbred rats reveal genetic regulatory mechanisms relevant for human obesity.

Transcriptomic analysis in metabolically active tissues allows a systems genetics approach to identify causal genes and networks involved in metabolic disease. Outbred heterogeneous stock (HS) rats are used for genetic mapping of complex traits, but to-date, a systems genetics analysis of metabolic tissues has not been done. We investigated whether adiposity-associated genes and gene coexpression networks in outbred heterogeneous stock (HS) rats overlap those found in humans. We analyzed RNAseq data from adipose tissue of 415 male HS rats, correlated these transcripts with body weight (BW) and compared transcriptome signatures to two human cohorts: the "African American Genetics of Metabolism and Expression" and "Metabolic Syndrome in Men." We used weighted gene coexpression network analysis to identify adiposity-associated gene networks and mediation analysis to identify genes under genetic control whose expression drives adiposity. We identified 554 orthologous "consensus genes" whose expression correlates with BW in the rat and with body mass index (BMI) in both human cohorts. Consensus genes fell within eight coexpressed networks and were enriched for genes involved in immune system function, cell growth, extracellular matrix organization, and lipid metabolic processes. We identified 19 consensus genes for which genetic variation may influence BW via their expression, including those involved in lipolysis (e.g., Hcar1), inflammation (e.g., Rgs1), adipogenesis (e.g., Tmem120b), or no previously known role in obesity (e.g., St14 and Ms4a6a). Strong concordance between HS rat and human BW/BMI associated transcripts demonstrates translational utility of the rat model, while identification of novel genes expands our knowledge of the genetics underlying obesity.

Integrative analysis reveals mouse strain-dependent responses to acute ozone exposure associated with airway macrophage transcriptional activity.

Acute ozone (O3) exposure is associated with multiple adverse cardiorespiratory outcomes, the severity of which varies across individuals in human populations and inbred mouse strains. However, molecular determinants of response, including susceptibility biomarkers that distinguish who will develop severe injury and inflammation, are not well characterized. We and others have demonstrated that airway macrophages (AMs) are an important resident immune cell type that are functionally and transcriptionally responsive to O3 inhalation. Here, we sought to explore influences of strain, exposure, and strain-by-O3 exposure interactions on AM gene expression and identify transcriptional correlates of O3-induced inflammation and injury across six mouse strains, including five Collaborative Cross (CC) strains. We exposed adult mice of both sexes to filtered air (FA) or 2 ppm O3 for 3 h and measured inflammatory and injury parameters 21 h later. Mice exposed to O3 developed airway neutrophilia and lung injury with strain-dependent severity. In AMs, we identified a common core O3 transcriptional response signature across all strains, as well as a set of genes exhibiting strain-by-O3 exposure interactions. In particular, a prominent gene expression contrast emerged between a low- (CC017/Unc) and high-responding (CC003/Unc) strain, as reflected by cellular inflammation and injury. Further inspection indicated that differences in their baseline gene expression and chromatin accessibility profiles likely contribute to their divergent post-O3 exposure transcriptional responses. Together, these results suggest that aspects of O3-induced respiratory responses are mediated through altered AM transcriptional signatures and further confirm the importance of gene-environment interactions in mediating differential responsiveness to environmental agents.

Inferring the Allelic Series at QTL in Multiparental Populations.

Multiparental populations (MPPs) are experimental populations in which the genome of every individual is a mosaic of known founder haplotypes. These populations are useful for detecting quantitative trait loci (QTL) because tests of association can leverage inferred founder haplotype descent. It is difficult, however, to determine how haplotypes at a locus group into distinct functional alleles, termed the allelic series. The allelic series is important because it provides information about the number of causal variants at a QTL and their combined effects. In this study, we introduce a fully Bayesian model selection framework for inferring the allelic series. This framework accounts for sources of uncertainty found in typical MPPs, including the number and composition of functional alleles. Our prior distribution for the allelic series is based on the Chinese restaurant process, a relative of the Dirichlet process, and we leverage its connection to the coalescent to introduce additional prior information about haplotype relatedness via a phylogenetic tree. We evaluate our approach via simulation and apply it to QTL from two MPPs: the Collaborative Cross (CC) and the Drosophila Synthetic Population Resource (DSPR). We find that, although posterior inference of the exact allelic series is often uncertain, we are able to distinguish biallelic QTL from more complex multiallelic cases. Additionally, our allele-based approach improves haplotype effect estimation when the true number of functional alleles is small. Our method, Tree-Based Inference of Multiallelism via Bayesian Regression (TIMBR), provides new insight into the genetic architecture of QTL in MPPs.

Transcriptional Profiling of the Murine Airway Response to Acute Ozone Exposure.

Ambient ozone (O3) exposure has serious consequences on respiratory health, including airway inflammation and injury. Decades of research have yielded thorough descriptions of these outcomes; however, less is known about the molecular processes that drive them. The aim of this study was to further describe the cellular and molecular responses to O3 exposure in murine airways, with a particular focus on transcriptional responses in 2 critical pulmonary tissue compartments: conducting airways (CA) and airway macrophages (AM). After exposing adult, female C57BL/6J mice to filtered air, 1 or 2 ppm O3, we assessed hallmark responses including airway inflammation (cell counts and cytokine secretion) and injury (epithelial permeability), followed by gene expression profiling of CA and AM by RNA-seq. As expected, we observed concentration-dependent increases in airway inflammation and injury. Conducting airways and AM both exhibited changes in gene expression to both 1 and 2 ppm O3 that were largely compartment-specific. In CA, genes associated with epithelial barrier function, detoxification processes, and cellular proliferation were altered, while O3 affected genes involved in innate immune signaling, cytokine production, and extracellular matrix remodeling in AM. Further, CA and AM also exhibited notable differences in concentration-response expression patterns for large numbers of genes. Overall, our study has described transcriptional responses to acute O3 exposure, revealing both shared and unique gene expression patterns across multiple concentrations of O3 and in 2 important O3-responsive tissues. These profiles provide broad mechanistic insight into pulmonary O3 toxicity, and reveal a variety of targets for focused follow-up studies.

Determinants of QTL Mapping Power in the Realized Collaborative Cross.

The Collaborative Cross (CC) is a mouse genetic reference population whose range of applications includes quantitative trait loci (QTL) mapping. The design of a CC QTL mapping study involves multiple decisions, including which and how many strains to use, and how many replicates per strain to phenotype, all viewed within the context of hypothesized QTL architecture. Until now, these decisions have been informed largely by early power analyses that were based on simulated, hypothetical CC genomes. Now that more than 50 CC strains are available and more than 70 CC genomes have been observed, it is possible to characterize power based on realized CC genomes. We report power analyses from extensive simulations and examine several key considerations: 1) the number of strains and biological replicates, 2) the QTL effect size, 3) the presence of population structure, and 4) the distribution of functionally distinct alleles among the founder strains at the QTL. We also provide general power estimates to aide in the design of future experiments. All analyses were conducted with our R package, SPARCC (Simulated Power Analysis in the Realized Collaborative Cross), developed for performing either large scale power analyses or those tailored to particular CC experiments.

Explaining variation across grantees in breast and cervical cancer screening proportions in the NBCCEDP.

PURPOSE: There is substantial variation across the National Breast and Cervical Cancer Early Detection Program (NBCCEDP) grantees in terms of the proportion of the eligible population served by the grantees each year (hereafter referred to as the screening proportion). In this paper, we assess program- and state-level factors to better understand the reason for this variation in breast and cervical cancer screening proportions across the NBCCEDP grantees. METHODS: We constructed a longitudinal data set, consisting of data from NBCCEDP grantees for each of the three study years (program-years 2006-2007, 2008-2009, and 2009-2010). We performed multivariate analysis to explain the variation in breast and cervical cancer screening proportions across the grantees. The program-level factors studied were the total federal funds received, average cost of screening women by grantee, and the overall organizational structure. The state-level variables included were urban versus rural mix, access to care, and the size of the eligible population. RESULTS: Of the 48 grantees included in the study, those that serve larger populations, as measured by the size of the population and the percentage of women eligible for services, had lower screening proportions. Higher average cost of service delivery was also associated with lower screening proportions. In addition, grantees whose populations were more concentrated in urban areas had lower screening proportions. CONCLUSIONS: Overall, the average cost of screening, the overall size of the population eligible, and the concentration of population in urban areas all had a negative relationship to the proportion of eligible women screened by NBCCEDP grantees.

Medicare's intensive behavioral therapy for obesity: an exploratory cost-effectiveness analysis.

INTRODUCTION: Medicare coverage recently was expanded to include intensive behavioral therapy for obese individuals in primary care settings. PURPOSE: To examine the potential cost effectiveness of Medicare's intensive behavioral therapy for obesity, accounting for uncertainty in effectiveness and utilization. METHODS: A Markov simulation model of type 2 diabetes was used to estimate long-term health benefits and healthcare system costs of intensive behavioral therapy for obesity in the Medicare population without diabetes relative to an alternative of usual care. Cohort statistics were based on the 2005-2008 National Health and Nutrition Examination Survey. Model parameters were derived from the literature. Analyses were conducted in 2014 and reported in 2012 U.S. dollars. RESULTS: Based on assumptions for the maximal intervention effectiveness, intensive behavioral therapy is likely to be cost saving if costs per session equal the current reimbursement rate ($25.19) and will provide a cost-effectiveness ratio of $20,912 per quality-adjusted life-year if costs equal the rate for routine office visits. The intervention is less cost effective if it is less effective in primary care settings or if fewer intervention sessions are supplied by providers or used by participants. CONCLUSIONS: If the effectiveness of the intervention is similar to lifestyle interventions tested in other settings and costs per session equal the current reimbursement rate, intensive behavioral therapy for obesity offers good value. However, intervention effectiveness and the pattern of implementation and utilization strongly influence cost effectiveness. Given uncertainty regarding these factors, additional data might be collected to validate the modeling results.

Cost of services provided by the National Breast and Cervical Cancer Early Detection Program.

BACKGROUND: The National Breast and Cervical Cancer Early Detection Program (NBCCEDP) is the largest cancer screening program for low-income women in the United States. This study updates previous estimates of the costs of delivering preventive cancer screening services in the NBCCEDP. METHODS: We developed a standardized web-based cost-assessment tool to collect annual activity-based cost data on screening for breast and cervical cancer in the NBCCEDP. Data were collected from 63 of the 66 programs that received funding from the Centers for Disease Control and Prevention during the 2006/2007 fiscal year. We used these data to calculate costs of delivering preventive public health services in the program. RESULTS: We estimated the total cost of all NBCCEDP services to be $296 (standard deviation [SD], $123) per woman served (including the estimated value of in-kind donations, which constituted approximately 15% of this total estimated cost). The estimated cost of screening and diagnostic services was $145 (SD, $38) per women served, which represented 57.7% of the total cost excluding the value of in-kind donations. Including the value of in-kind donations, the weighted mean cost of screening a woman for breast cancer was $110 with an office visit and $88 without, the weighted mean cost of a diagnostic procedure was $401, and the weighted mean cost per breast cancer detected was $35,480. For cervical cancer, the corresponding cost estimates were $61, $21, $415, and $18,995, respectively. CONCLUSIONS: These NBCCEDP cost estimates may help policy makers in planning and implementing future costs for various potential changes to the program.

Economies of scale in federally-funded state-organized public health programs: results from the National Breast and Cervical Cancer Early Detection Programs.

This study investigates the existence of economies of scale in the provision of breast and cervical cancer screening and diagnostic services by state National Breast and Cervical Cancer Early Detection Program (NBCCEDP) grantees. A translog cost function is estimated as a system with input factor share equations. The estimated cost function is then used to determine output levels for which average costs are decreasing (i.e., economies of scale exist). Data were collected from all state NBCCEDP programs and District of Columbia for program years 2006-2007, 2008-2009 and 2009-2010 (N = 147). Costs included all programmatic and in-kind contributions from federal and non-federal sources, allocated to breast and cervical cancer screening activities. Output was measured by women served, women screened and cancers detected, separately by breast and cervical services for each measure. Inputs included labor, rent and utilities, clinical services, and quasi-fixed factors (e.g., percent of women eligible for screening by the NBCCEDP). 144 out of 147 program-years demonstrated significant economies of scale for women served and women screened; 136 out of 145 program-years displayed significant economies of scale for cancers detected. The cost data were self-reported by the NBCCEDP State programs. Quasi-fixed inputs were allowed to affect costs but not economies of scale or the share equations. The main analysis accounted for clustering of observations within State programs, but it did not make full use of the panel data. The average cost of providing breast and cervical cancer screening services decreases as the number of women screened and served increases.

The economic burden of vision loss and eye disorders among the United States population younger than 40 years.

OBJECTIVE: To estimate the economic burden of vision loss and eye disorders in the United States population younger than 40 years in 2012. DESIGN: Econometric and statistical analysis of survey, commercial claims, and census data. PARTICIPANTS: The United States population younger than 40 years in 2012. METHODS: We categorized costs based on consensus guidelines. We estimated medical costs attributable to diagnosed eye-related disorders, undiagnosed vision loss, and medical vision aids using Medical Expenditure Panel Survey and MarketScan data. The prevalence of vision impairment and blindness were estimated using National Health and Nutrition Examination Survey data. We estimated costs from lost productivity using Survey of Income and Program Participation. We estimated costs of informal care, low vision aids, special education, school screening, government spending, and transfer payments based on published estimates and federal budgets. We estimated quality-adjusted life years (QALYs) lost based on published utility values. MAIN OUTCOME MEASURES: Costs and QALYs lost in 2012. RESULTS: The economic burden of vision loss and eye disorders among the United States population younger than 40 years was $27.5 billion in 2012 (95% confidence interval, $21.5-$37.2 billion), including $5.9 billion for children and $21.6 billion for adults 18 to 39 years of age. Direct costs were $14.5 billion, including $7.3 billion in medical costs for diagnosed disorders, $4.9 billion in refraction correction, $0.5 billion in medical costs for undiagnosed vision loss, and $1.8 billion in other direct costs. Indirect costs were $13 billion, primarily because of $12.2 billion in productivity losses. In addition, vision loss cost society 215 000 QALYs. CONCLUSIONS: We found a substantial burden resulting from vision loss and eye disorders in the United States population younger than 40 years, a population excluded from previous studies. Monetizing quality-of-life losses at $50 000 per QALY would add $10.8 billion in additional costs, indicating a total economic burden of $38.2 billion. Relative to previously reported estimates for the population 40 years of age and older, more than one third of the total cost of vision loss and eye disorders may be incurred by persons younger than 40 years. FINANCIAL DISCLOSURE(S): The author(s) have no proprietary or commercial interest in any materials discussed in this article.