A unifying framework for evaluating the predictive power of genetic variants based on the level of heritability explained

ADuLT: An efficient and robust time-to-event GWAS

Proportional hazards models have been proposed to analyse time-to-event phenotypes in genome-wide association studies (GWAS). However, little is known about the ability of proportional hazards models to identify genetic associations under different generative models and when ascertainment is present. Here we propose the age-dependent liability threshold (ADuLT) model as an alternative to a Cox regression based GWAS, here represented by SPACox. We compare ADuLT, SPACox, and standard case-control GWAS in simulations under two generative models and with varying degrees of ascertainment as well as in the iPSYCH cohort. We find Cox regression GWAS to be underpowered when cases are strongly ascertained (cases are oversampled by a factor 5), regardless of the generative model used. ADuLT is robust to ascertainment in all simulated scenarios. Then, we analyse four psychiatric disorders in iPSYCH, ADHD, Autism, Depression, and Schizophrenia, with a strong case-ascertainment. Across these psychiatric disorders, ADuLT identifies 20 independent genome-wide significant associations, case-control GWAS finds 17, and SPACox finds 8, which is consistent with simulation results. As more genetic data are being linked to electronic health records, robust GWAS methods that can make use of age-of-onset information will help increase power in analyses for common health outcomes.

Accounting for age of onset and family history improves power in genome-wide association studies

Genome-wide association studies (GWASs) have revolutionized human genetics, allowing researchers to identify thousands of disease-related genes and possible drug targets. However, case-control status does not account for the fact that not all controls may have lived through their period of risk for the disorder of interest. This can be quantified by examining the age-of-onset distribution and the age of the controls or the age of onset for cases. The age-of-onset distribution may also depend on information such as sex and birth year. In addition, family history is not routinely included in the assessment of control status. Here, we present LT-FH++, an extension of the liability threshold model conditioned on family history (LT-FH), which jointly accounts for age of onset and sex as well as family history. Using simulations, we show that, when family history and the age-of-onset distribution are available, the proposed approach yields statistically significant power gains over LT-FH and large power gains over genome-wide association study by proxy (GWAX). We applied our method to four psychiatric disorders available in the iPSYCH data and to mortality in the UK Biobank and found 20 genome-wide significant associations with LT-FH++, compared to ten for LT-FH and eight for a standard case-control GWAS. As more genetic data with linked electronic health records become available to researchers, we expect methods that account for additional health information, such as LT-FH++, to become even more beneficial.

Analysis of genetic differences between psychiatric disorders: exploring pathways and cell types/tissues involved and ability to differentiate the disorders by polygenic scores

Although displaying genetic correlations, psychiatric disorders are clinically defined as categorical entities as they each have distinguishing clinical features and may involve different treatments. Identifying differential genetic variations between these disorders may reveal how the disorders differ biologically and help to guide more personalized treatment. Here we presented a statistical framework and comprehensive analysis to identify genetic markers differentially associated with various psychiatric disorders/traits based on GWAS summary statistics, covering 18 psychiatric traits/disorders and 26 comparisons. We also conducted comprehensive analysis to unravel the genes, pathways and SNP functional categories involved, and the cell types and tissues implicated. We also assessed how well one could distinguish between psychiatric disorders by polygenic risk scores (PRS). SNP-based heritabilities (h2snp) were significantly larger than zero for most comparisons. Based on current GWAS data, PRS have mostly modest power to distinguish between psychiatric disorders. For example, we estimated that AUC for distinguishing schizophrenia from major depressive disorder (MDD), bipolar disorder (BPD) from MDD and schizophrenia from BPD were 0.694, 0.602 and 0.618, respectively, while the maximum AUC (based on h2snp) were 0.763, 0.749 and 0.726, respectively. We also uncovered differences in each pair of studied traits in terms of their differences in genetic correlation with comorbid traits. For example, clinically defined MDD appeared to more strongly genetically correlated with other psychiatric disorders and heart disease, when compared to non-clinically defined depression in UK Biobank. Our findings highlight genetic differences between psychiatric disorders and the mechanisms involved. PRS may help differential diagnosis of selected psychiatric disorders in the future with larger GWAS samples.

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.

Prediction of Early Childhood Caries Based on Single Nucleotide Polymorphisms Using Neural Networks

Background: Several genes and single nucleotide polymorphisms (SNPs) have been associated with early childhood caries. However, they are highly age- and population-dependent and the majority of existing caries prediction models are based on environmental and behavioral factors only and are scarce in infants. Methods: We examined 6 novel and previously analyzed 22 SNPs in the cohort of 95 Polish children (48 caries, 47 caries-free) aged 2–3 years. All polymorphisms were genotyped from DNA extracted from oral epithelium samples. We used Fisher’s exact test, receiver operator characteristic (ROC) curve and uni-/multi-variable logistic regression to test the association of SNPs with the disease, followed by the neural network (NN) analysis. Results: The logistic regression (LogReg) model showed 90% sensitivity and 96% specificity, overall accuracy of 93% (p < 0.0001), and the area under the curve (AUC) was 0.970 (95% CI: 0.912–0.994; p < 0.0001). We found 90.9–98.4% and 73.6–87.2% prediction accuracy in the test and validation predictions, respectively. The strongest predictors were: AMELX_rs17878486 and TUFT1_rs2337360 (in both LogReg and NN), MMP16_rs1042937 (in NN) and ENAM_rs12640848 (in LogReg). Conclusions: Neural network prediction model might be a substantial tool for screening/early preventive treatment of patients at high risk of caries development in the early childhood. The knowledge of potential risk status could allow early targeted training in oral hygiene and modifications of eating habits.

Potential Distribution of the Biocontrol Agent Toxorhynchites rutilus By 2070

Climate change projections indicate that mosquito distributions will expand to include new areas of North America, increasing human exposure to mosquito-borne disease. Controlling these vectors is imperative, as mosquito-borne disease incidence will rise in response to expansion of mosquito range and increased seasonality. One means of mosquito control used in the USA is the biocontrol agent, Toxorhynchites rutilus. Climate change will open new habitats for its use by vector control organizations, but the extent of this change in habitat is currently unknown. We used a maximum entropy approach to create species distribution models for Tx. rutilus under 4 climate change scenarios by 2070. Mean temperature of warmest quarter (22.6°C to 29.1°C), annual precipitation (1,025.15 mm to 1,529.40 mm), and precipitation seasonality (≤17.86) are the most important bioclimatic variables for suitable habitat. The center of current possible habitat distribution of Tx. rutilus is in central Tennessee. Depending upon the scenario, we expect centroids to shift north-northeast by 97.68 km to 280.16 km by 2070. The extreme change in area of greater than 50% suitable habitat probability is 141.14% with 99.44% area retained. Our models indicate limited change in current habitat as well as creation of new habitat. These results are promising for North American mosquito control programs for the continued and potential combat of vector mosquitoes using Tx. rutilus.

Criteria for evaluating risk prediction of multiple outcomes

Risk prediction models have been developed in many contexts to classify individuals according to a single outcome, such as risk of a disease. Emerging “-omic” biomarkers provide panels of features that can simultaneously predict multiple outcomes from a single biological sample, creating issues of multiplicity reminiscent of exploratory hypothesis testing. Here I propose definitions of some basic criteria for evaluating prediction models of multiple outcomes. I define calibration in the multivariate setting and then distinguish between outcome-wise and individual-wise prediction, and within the latter between joint and panel-wise prediction. I give examples such as screening and early detection in which different senses of prediction may be more appropriate. In each case I propose definitions of sensitivity, specificity, concordance, positive and negative predictive value and relative utility. I link the definitions through a multivariate probit model, showing that the accuracy of a multivariate prediction model can be summarised by its covariance with a liability vector. I illustrate the concepts on a biomarker panel for early detection of eight cancers, and on polygenic risk scores for six common diseases.

Towards clinical utility of polygenic risk scores

Prediction of disease risk is an essential part of preventative medicine, often guiding clinical management. Risk prediction typically includes risk factors such as age, sex, family history of disease and lifestyle (e.g. smoking status); however, in recent years, there has been increasing interest to include genomic information into risk models. Polygenic risk scores (PRS) aggregate the effects of many genetic variants across the human genome into a single score and have recently been shown to have predictive value for multiple common diseases. In this review, we summarize the potential use cases for seven common diseases (breast cancer, prostate cancer, coronary artery disease, obesity, type 1 diabetes, type 2 diabetes and Alzheimer’s disease) where PRS has or could have clinical utility. PRS analysis for these diseases frequently revolved around (i) risk prediction performance of a PRS alone and in combination with other non-genetic risk factors, (ii) estimation of lifetime risk trajectories, (iii) the independent information of PRS and family history of disease or monogenic mutations and (iv) estimation of the value of adding a PRS to specific clinical risk prediction scenarios. We summarize open questions regarding PRS usability, ancestry bias and transferability, emphasizing the need for the next wave of studies to focus on the implementation and health-economic value of PRS testing. In conclusion, it is becoming clear that PRS have value in disease risk prediction and there are multiple areas where this may have clinical utility.

Predictive accuracy of combined genetic and environmental risk scores

The substantial heritability of most complex diseases suggests that genetic data could provide useful risk prediction. To date the performance of genetic risk scores has fallen short of the potential implied by heritability, but this can be explained by insufficient sample sizes for estimating highly polygenic models. When risk predictors already exist based on environment or lifestyle, two key questions are to what extent can they be improved by adding genetic information, and what is the ultimate potential of combined genetic and environmental risk scores? Here, we extend previous work on the predictive accuracy of polygenic scores to allow for an environmental score that may be correlated with the polygenic score, for example when the environmental factors mediate the genetic risk. We derive common measures of predictive accuracy and improvement as functions of the training sample size, chip heritabilities of disease and environmental score, and genetic correlation between disease and environmental risk factors. We consider simple addition of the two scores and a weighted sum that accounts for their correlation. Using examples from studies of cardiovascular disease and breast cancer, we show that improvements in discrimination are generally small but reasonable degrees of reclassification could be obtained with current sample sizes. Correlation between genetic and environmental scores has only minor effects on numerical results in realistic scenarios. In the longer term, as the accuracy of polygenic scores improves they will come to dominate the predictive accuracy compared to environmental scores.

Exploring the predictive power of polygenic scores derived from genome-wide association studies: a study of 10 complex traits

Motivation

It is hoped that advances in our knowledge in disease genomics will contribute to personalized medicine such as individualized preventive strategies or early diagnoses of diseases. With the growth of genome-wide association studies (GWAS) in the past decade, how far have we reached this goal? In this study we explored the predictive ability of polygenic risk scores (PRSs) derived from GWAS for a range of complex disease and traits.

Results

We first proposed a new approach to evaluate predictive performances of PRS at arbitrary P -value thresholds. The method was based on corrected estimates of effect sizes, accounting for possible false positives and selection bias. This approach requires no distributional assumptions and only requires summary statistics as input. The validity of the approach was verified in simulations. We explored the predictive power of PRS for ten complex traits, including type 2 diabetes (DM), coronary artery disease (CAD), triglycerides, high- and low-density lipoprotein, total cholesterol, schizophrenia (SCZ), bipolar disorder (BD), major depressive disorder and anxiety disorders. We found that the predictive ability of PRS for CAD and DM were modest (best AUC = 0.608 and 0.607) while for lipid traits the prediction R-squared ranged from 16.1 to 29.8%. For psychiatric disorders, the predictive power for SCZ was estimated to be the highest (best AUC 0.820), followed by BD. Predictive performance of other psychiatric disorders ranged from 0.543 to 0.585. Psychiatric traits tend to have more gradual rise in AUC when significance thresholds increase and achieve the best predictive power at higher P -values than cardiometabolic traits.

Contact

hcso@cuhk.edu.hk ; pcsham@hku.hk.

Supplementary information

Supplementary data are available at Bioinformatics online.

Asymptotic distribution of ∆AUC, NRIs, and IDI based on theory of U-statistics

The change in area under the curve (∆AUC), the integrated discrimination improvement (IDI), and net reclassification index (NRI) are commonly used measures of risk prediction model performance. Some authors have reported good validity of associated methods of estimating their standard errors (SE) and construction of confidence intervals, whereas others have questioned their performance. To address these issues, we unite the ∆AUC, IDI, and three versions of the NRI under the umbrella of the U-statistics family. We rigorously show that the asymptotic behavior of ∆AUC, NRIs, and IDI fits the asymptotic distribution theory developed for U-statistics. We prove that the ∆AUC, NRIs, and IDI are asymptotically normal, unless they compare nested models under the null hypothesis. In the latter case, asymptotic normality and existing SE estimates cannot be applied to ∆AUC, NRIs, or IDI. In the former case, SE formulas proposed in the literature are equivalent to SE formulas obtained from U-statistics theory if we ignore adjustment for estimated parameters. We use Sukhatme-Randles-deWet condition to determine when adjustment for estimated parameters is necessary. We show that adjustment is not necessary for SEs of the ∆AUC and two versions of the NRI when added predictor variables are significant and normally distributed. The SEs of the IDI and three-category NRI should always be adjusted for estimated parameters. These results allow us to define when existing formulas for SE estimates can be used and when resampling methods such as the bootstrap should be used instead when comparing nested models. We also use the U-statistic theory to develop a new SE estimate of ∆AUC. Copyright © 2017 John Wiley & Sons, Ltd.

Projected changes in climatic suitability for Kinosternon turtles by 2050 and 2070

Chelonians are expected to be negatively impacted by climate change due to limited vagility and temperature‐dependent sex determination. However, few studies have examined how freshwater turtle distributions may shift under different climate change scenarios. We used a maximum entropy approach to model the distribution of five widespread North American Kinosternon species (K. baurii, K. flavescens, K. hirtipes, K. sonoriense, and K. subrubrum) under four climate change scenarios. We found that areas with suitable climatic conditions for K. baurii and K. hirtipes are expected to decline substantially during the 21st century. In contrast, the area with suitable climate for K. sonoriense will remain essentially unchanged, while areas suitable for K. flavescens and K. subrubrum are expected to substantially increase. The centroid for the distribution of four of the five species shifted northward, while the centroid for K. sonoriense shifted slightly southward. Overall, centroids shifted at a median rate of 37.5 km per decade across all scenarios. Given the limited dispersal ability of turtles, it appears unlikely that range shifts will occur rapidly enough to keep pace with climate change during the 21st century. The ability of chelonians to modify behavioral and physiological responses in response to unfavorable conditions may allow turtles to persist for a time in areas that have become increasingly unsuitable, but this plasticity will likely only delay local extinctions.

Validity of using ad hoc methods to analyze secondary traits in case-control association studies

Case-control association studies often collect from their subjects information on secondary phenotypes. Reusing the data and studying the association between genes and secondary phenotypes provide an attractive and cost-effective approach that can lead to discovery of new genetic associations. A number of approaches have been proposed, including simple and computationally efficient ad hoc methods that ignore ascertainment or stratify on case-control status. Justification for these approaches relies on the assumption of no covariates and the correct specification of the primary disease model as a logistic model. Both might not be true in practice, for example, in the presence of population stratification or the primary disease model following a probit model. In this paper, we investigate the validity of ad hoc methods in the presence of covariates and possible disease model misspecification. We show that in taking an ad hoc approach, it may be desirable to include covariates that affect the primary disease in the secondary phenotype model, even though these covariates are not necessarily associated with the secondary phenotype. We also show that when the disease is rare, ad hoc methods can lead to severely biased estimation and inference if the true disease model follows a probit model instead of a logistic model. Our results are justified theoretically and via simulations. Applied to real data analysis of genetic associations with cigarette smoking, ad hoc methods collectively identified as highly significant (P<10-5) single nucleotide polymorphisms from over 10 genes, genes that were identified in previous studies of smoking cessation.

Development of risk prediction models for glioma based on genome-wide association study findings and comprehensive evaluation of predictive performances

Over 14 common single nucleotide polymorphisms (SNP) have been consistently identified from genome-wide association studies (GWAS) as associated with glioma risk in European background. The extent to which and how these genetic variants can improve the prediction of glioma risk has was not been investigated. In this study, we employed three independent case-control datasets in Chinese populations, tested GWAS signals in dataset1, validated association results in dataset2, developed prediction models in dataset2 for the consistently replicated SNPs, refined the consistently replicated SNPs in dataset3 and developed tailored models for Chinese populations. For model construction, we aggregated the contribution of multiple SNPs into genetic risk scores (count GRS and weighed GRS) or predicted risks from logistic regression analyses (PRFLR).

In dataset2, the area under receiver operating characteristic curves (AUC) of the 5 consistently replicated SNPs by PRFLR(SNPs) was 0.615, higher than those of all GRSs(ranging from 0.607 to 0.611, all P>0.05). The AUC of genetic profile significantly exceeded that of family history (fmc) alone (AUC=0.535, all P<0.001). The best model in our study comprised “PRURA +fmc” (AUC=0.646) in dataset3. Further model assessment analyses provided additional evidence.

This study indicates that genetic markers have potential value for risk prediction of glioma.

What to expect from net reclassification improvement with three categories

The net reclassification improvement (NRI) has become a popular measure of incremental usefulness of markers added to risk prediction models. However, the expected magnitude of the three-category NRI is not well understood, leading researchers to rely on statistical significance. In this paper, we describe a slight modification to the original definition of the NRI, which weighs each reclassification by the number of categories by which a given individual is reclassified. This modification resolves some recent criticisms of the three-category NRI and at the same time has a minimal impact on its magnitude. Then we show that using this modified definition, the event and nonevent NRIs have simple interpretations as sums of changes in sensitivities and specificities calculated at the risk thresholds. We exploit this relationship to arrive at closed-form solutions for the NRI under normality within the event and nonevent subgroups. We observe that the size of the intermediate risk category and the event rate have limited impact on the magnitude of the NRI. As expected, the NRI increases with the strength of the added marker, and this relationship appears fairly proportional for markers with non-weak net effect size (above 0.25). Furthermore, we conclude that using the NRI as a metric, it is harder to improve models that already perform well.

Global decline in suitable habitat for Angiostrongylus ( = Parastrongylus) cantonensis: the role of climate change

Climate change is implicated in the alteration of the ranges of species worldwide. Such shifts in species distributions may introduce parasites/pathogens, hosts, and vectors associated with disease to new areas. The parasite Angiostrongylus ( = Parastrongylus) cantonensis is an invasive species that causes eosinophilic meningitis in humans and neurological abnormalities in domestic/wild animals. Although native to southeastern Asia, A. cantonensis has now been reported from more than 30 countries worldwide. Given the health risks, it is important to describe areas with potentially favorable climate for the establishment of A. cantonensis, as well as areas where this pathogen might become established in the future. We used the program Maxent to develop an ecological niche model for A. cantonensis based on 86 localities obtained from published literature. We then modeled areas of potential A. cantonensis distribution as well as areas projected to have suitable climatic conditions under four Representative Concentration Pathways (RCP) scenarios by the 2050s and the 2070s. The best model contained three bioclimatic variables: mean diurnal temperature range, minimum temperature of coldest month and precipitation of warmest quarter. Potentially suitable habitat for A. cantonensis was located worldwide in tropical and subtropical regions. Under all climate change RCP scenarios, the center of the projected distribution shifted away from the equator at a rate of 68–152 km per decade. However, the extent of areas with highly suitable habitat (>50%) declined by 10.66–15.66% by the 2050s and 13.11–16.11% by the 2070s. These results conflict with previous studies, which have generally found that the prevalence of tropical pathogens will increase during the 21st century. Moreover, it is likely that A. cantonensis will continue to expand its current range in the near future due to introductions and host expansion, whereas climate change will reduce the total geographic area of most suitable climatic conditions during the coming decades.

The emerging molecular architecture of schizophrenia, polygenic risk scores and the clinical implications for GxE research

Schizophrenia is a devastating mental disorder. The level of risk in the general population is sustained by the persistence of social, environmental and biological factors, as well as their interactions. Socio-environmental risk factors for schizophrenia are well established and robust. The same can belatedly be said of genetic risk factors for the disorder. Recent progress in schizophrenia genetics is primarily fuelled by genome-wide association, which is able to leverage substantial proportions of additional explained variance previously classified as 'missing'. Here, we provide an outline of the emerging genetic landscape of schizophrenia and demonstrate how this knowledge can be turned into a simple empirical measure of genetic risk, known as a polygenic risk score. We highlight the statistical framework used to assess the clinical potential of the new score and finally, draw relevance to and discuss the clinical implications for the study of gene-environment interaction.

Common variants explain a large fraction of the variability in the liability to psoriasis in a Han Chinese population

Background

Psoriasis is a common inflammatory skin disease with a known genetic component. Our previously published psoriasis genome-wide association study identified dozens of novel susceptibility loci in Han Chinese. However, these markers explained only a small fraction of the estimated heritable component of psoriasis. To better understand the unknown yet likely polygenic architecture in psoriasis, we applied a linear mixed model to quantify the variation in the liability to psoriasis explained by common genetic markers (minor allele frequency > 0.01) in a Han Chinese population.

Results

We explored the polygenic genetic architecture of psoriasis using genome-wide association data from 2,271 Han Chinese individuals. We estimated that 34.9% (s.e. = 6.0%, P = 9 × 10^-9) of the variation in the liability to psoriasis is captured by common genotyped and imputed variants. We discuss these results in the context of the strong association between HLA variants and psoriasis. We also show that the variance explained by each chromosome is linearly correlated to its length (R² = 0.27, P=0.01), and quantify the impact of a polygenic effect on the prediction and diagnosis of psoriasis.

Conclusions

Our results suggest that psoriasis has a substantial polygenic component, which not only has implications for the development of genetic diagnostics and prognostics for psoriasis, but also suggests that more individual variants contributing to psoriasis may be detected if sample sizes in future association studies are increased.

Role of Environmental Confounding in the Association between FKBP5 and First-Episode Psychosis

BACKGROUND

Failure to account for the etiological diversity that typically occurs in psychiatric cohorts may increase the potential for confounding as a proportion of genetic variance will be specific to exposures that have varying distributions in cases. This study investigated whether minimizing the potential for such confounding strengthened the evidence for a genetic candidate currently unsupported at the genome-wide level.

METHODS

Two hundred and ninety-one first-episode psychosis cases from South London, UK and 218 unaffected controls were evaluated for a functional polymorphism at the rs1360780 locus in FKBP5. The relationship between FKBP5 and psychosis was modeled using logistic regression. Cannabis use (Cannabis Experiences Questionnaire) and parental separation (Childhood Experience of Care and Abuse Questionnaire) were included as confounders in the analysis.

RESULTS

Association at rs1360780 was not detected until the effects of the two environmental factors had been adjusted for in the model (OR = 2.81, 95% CI 1.23-6.43, p = 0.02). A statistical interaction between rs1360780 and parental separation was confirmed by stratified tests (OR = 2.8, p = 0.02 vs. OR = 0.89, p = 0.80). The genetic main effect was directionally consistent with findings in other (stress-related) clinical phenotypes. Moreover, the variation in effect magnitude was explained by the level of power associated with different cannabis constructs used in the model (r = 0.95).

CONCLUSION

Our results suggest that the extent to which genetic variants in FKBP5 can influence susceptibility to psychosis may depend on other etiological factors. This finding requires further validation in large independent cohorts. Potentially this work could have translational implications; the ability to discriminate between genetic etiologies based on a case-by-case understanding of previous environmental exposures would confer an important clinical advantage that would benefit the delivery of personalizable treatment strategies.

Novel genetic analysis for case-control genome-wide association studies: quantification of power and genomic prediction accuracy

Genome-wide association studies (GWAS) are routinely conducted for both quantitative and binary (disease) traits. We present two analytical tools for use in the experimental design of GWAS. Firstly, we present power calculations quantifying power in a unified framework for a range of scenarios. In this context we consider the utility of quantitative scores (e.g. endophenotypes) that may be available on cases only or both cases and controls. Secondly, we consider, the accuracy of prediction of genetic risk from genome-wide SNPs and derive an expression for genomic prediction accuracy using a liability threshold model for disease traits in a case-control design. The expected values based on our derived equations for both power and prediction accuracy agree well with observed estimates from simulations.

Power and predictive accuracy of polygenic risk scores

Polygenic scores have recently been used to summarise genetic effects among an ensemble of markers that do not individually achieve significance in a large-scale association study. Markers are selected using an initial training sample and used to construct a score in an independent replication sample by forming the weighted sum of associated alleles within each subject. Association between a trait and this composite score implies that a genetic signal is present among the selected markers, and the score can then be used for prediction of individual trait values. This approach has been used to obtain evidence of a genetic effect when no single markers are significant, to establish a common genetic basis for related disorders, and to construct risk prediction models. In some cases, however, the desired association or prediction has not been achieved. Here, the power and predictive accuracy of a polygenic score are derived from a quantitative genetics model as a function of the sizes of the two samples, explained genetic variance, selection thresholds for including a marker in the score, and methods for weighting effect sizes in the score. Expressions are derived for quantitative and discrete traits, the latter allowing for case/control sampling. A novel approach to estimating the variance explained by a marker panel is also proposed. It is shown that published studies with significant association of polygenic scores have been well powered, whereas those with negative results can be explained by low sample size. It is also shown that useful levels of prediction may only be approached when predictors are estimated from very large samples, up to an order of magnitude greater than currently available. Therefore, polygenic scores currently have more utility for association testing than predicting complex traits, but prediction will become more feasible as sample sizes continue to grow.

Recently there has been much interest in combining multiple genetic markers into a single score for predicting disease risk. Even if many of the individual markers have no detected effect, the combined score could be a strong predictor of disease. This has allowed researchers to demonstrate that some diseases have a strong genetic basis, even if few actual genes have been identified, and it has also revealed a common genetic basis for distinct diseases. These analyses have so far been performed opportunistically, with mixed results. Here I derive formulae based on the heritability of disease and size of the study, allowing researchers to plan their analyses from a more informed position. I show that discouraging results in some previous studies were due to the low number of subjects studied, but a modest increase in study size would allow more successful analysis. However, I also show that, for genetics to become useful for predicting individual risk of disease, hundreds of thousands of subjects may be needed to estimate the gene effects. This is larger than most existing studies, but will become more common in the near future, so that gene scores will become more useful for predicting disease than has appeared to date.

Informed conditioning on clinical covariates increases power in case-control association studies

Genetic case-control association studies often include data on clinical covariates, such as body mass index (BMI), smoking status, or age, that may modify the underlying genetic risk of case or control samples. For example, in type 2 diabetes, odds ratios for established variants estimated from low–BMI cases are larger than those estimated from high–BMI cases. An unanswered question is how to use this information to maximize statistical power in case-control studies that ascertain individuals on the basis of phenotype (case-control ascertainment) or phenotype and clinical covariates (case-control-covariate ascertainment). While current approaches improve power in studies with random ascertainment, they often lose power under case-control ascertainment and fail to capture available power increases under case-control-covariate ascertainment. We show that an informed conditioning approach, based on the liability threshold model with parameters informed by external epidemiological information, fully accounts for disease prevalence and non-random ascertainment of phenotype as well as covariates and provides a substantial increase in power while maintaining a properly controlled false-positive rate. Our method outperforms standard case-control association tests with or without covariates, tests of gene x covariate interaction, and previously proposed tests for dealing with covariates in ascertained data, with especially large improvements in the case of case-control-covariate ascertainment. We investigate empirical case-control studies of type 2 diabetes, prostate cancer, lung cancer, breast cancer, rheumatoid arthritis, age-related macular degeneration, and end-stage kidney disease over a total of 89,726 samples. In these datasets, informed conditioning outperforms logistic regression for 115 of the 157 known associated variants investigated (P-value = 1×10⁻⁹). The improvement varied across diseases with a 16% median increase in χ² test statistics and a commensurate increase in power. This suggests that applying our method to existing and future association studies of these diseases may identify novel disease loci.

This work describes a new methodology for analyzing genome-wide case-control association studies of diseases with strong correlations to clinical covariates, such as age in prostate cancer and body mass index in type 2 diabetes. Currently, researchers either ignore these clinical covariates or apply approaches that ignore the disease's prevalence and the study's ascertainment strategy. We take an alternative approach, leveraging external prevalence information from the epidemiological literature and constructing a statistic based on the classic liability threshold model of disease. Our approach not only improves the power of studies that ascertain individuals randomly or based on the disease phenotype, but also improves the power of studies that ascertain individuals based on both the disease phenotype and clinical covariates. We apply our statistic to seven datasets over six different diseases and a variety of clinical covariates. We found that there was a substantial improvement in test statistics relative to current approaches at known associated variants. This suggests that novel loci may be identified by applying our method to existing and future association studies of these diseases.

Trees Assembling Mann-Whitney approach for detecting genome-wide joint association among low-marginal-effect loci

Common complex diseases are likely influenced by the interplay of hundreds, or even thousands, of genetic variants. Converging evidence shows that genetic variants with low marginal effects (LMEs) play an important role in disease development. Despite their potential significance, discovering LME genetic variants and assessing their joint association on high-dimensional data (e.g., genome-wide data) remain a great challenge. To facilitate joint association analysis among a large ensemble of LME genetic variants, we proposed a computationally efficient and powerful approach, which we call Trees Assembling Mann-Whitney (TAMW). Through simulation studies and an empirical data application, we found that TAMW outperformed multifactor dimensionality reduction (MDR) and the likelihood ratio-based Mann-Whitney approach (LRMW) when the underlying complex disease involves multiple LME loci and their interactions. For instance, in a simulation with 20 interacting LME loci, TAMW attained a higher power (power = 0.931) than both MDR (power = 0.599) and LRMW (power = 0.704). In an empirical study of 29 known Crohn's disease (CD) loci, TAMW also identified a stronger joint association with CD than those detected by MDR and LRMW. Finally, we applied TAMW to Wellcome Trust CD GWAS to conduct a genome-wide analysis. The analysis of 459K single nucleotide polymorphisms was completed in 40 hrs using parallel computing, and revealed a joint association predisposing to CD (P-value = 2.763 × 10(-19)). Further analysis of the newly discovered association suggested that 13 genes, such as ATG16L1 and LACC1, may play an important role in CD pathophysiological and etiological processes.

Comparison of family history and SNPs for predicting risk of complex disease

The clinical utility of family history and genetic tests is generally well understood for simple Mendelian disorders and rare subforms of complex diseases that are directly attributable to highly penetrant genetic variants. However, little is presently known regarding the performance of these methods in situations where disease susceptibility depends on the cumulative contribution of multiple genetic factors of moderate or low penetrance. Using quantitative genetic theory, we develop a model for studying the predictive ability of family history and single nucleotide polymorphism (SNP)–based methods for assessing risk of polygenic disorders. We show that family history is most useful for highly common, heritable conditions (e.g., coronary artery disease), where it explains roughly 20%–30% of disease heritability, on par with the most successful SNP models based on associations discovered to date. In contrast, we find that for diseases of moderate or low frequency (e.g., Crohn disease) family history accounts for less than 4% of disease heritability, substantially lagging behind SNPs in almost all cases. These results indicate that, for a broad range of diseases, already identified SNP associations may be better predictors of risk than their family history–based counterparts, despite the large fraction of missing heritability that remains to be explained. Our model illustrates the difficulty of using either family history or SNPs for standalone disease prediction. On the other hand, we show that, unlike family history, SNP–based tests can reveal extreme likelihood ratios for a relatively large percentage of individuals, thus providing potentially valuable adjunctive evidence in a differential diagnosis.

In clinical practice, obtaining a detailed family history is often considered the standard-of-care for characterizing the inherited component of an individual's disease risk. Recently, genetic risk assessments based on the cumulative effect of known single nucleotide polymorphism (SNP) disease associations have been proposed as another potentially useful source of information. To date, however, little is known regarding the predictive power of each approach. In this study, we develop models based on quantitative genetic theory to analyze and compare family history and SNP–based models. Our models explain the impact of disease frequency and heritability on performance for each method, and reveal a wide range of scenarios (16 out of the 23 diseases considered) where SNP associations may already be better predictors of risk than family history. Our results confirm the difficulty of obtaining accurate prediction when SNP or family history–based methods are used alone, and they show the benefits of combining information from the two approaches. They also suggest that, in some situations, SNP associations may be potentially useful as supporting evidence alongside other types of clinical information. To our knowledge, this study is the first broad comparison of family history– and SNP–based methods across a wide range of health conditions.

How accurate can genetic predictions be?

Background

Pre-symptomatic prediction of disease and drug response based on genetic testing is a critical component of personalized medicine. Previous work has demonstrated that the predictive capacity of genetic testing is constrained by the heritability and prevalence of the tested trait, although these constraints have only been approximated under the assumption of a normally distributed genetic risk distribution.

Results

Here, we mathematically derive the absolute limits that these factors impose on test accuracy in the absence of any distributional assumptions on risk. We present these limits in terms of the best-case receiver-operating characteristic (ROC) curve, consisting of the best-case test sensitivities and specificities, and the AUC (area under the curve) measure of accuracy. We apply our method to genetic prediction of type 2 diabetes and breast cancer, and we additionally show the best possible accuracy that can be obtained from integrated predictors, which can incorporate non-genetic features.

Conclusion

Knowledge of such limits is valuable in understanding the implications of genetic testing even before additional associations are identified.

Inclusion of gene-gene and gene-environment interactions unlikely to dramatically improve risk prediction for complex diseases

Genome-wide association studies have identified hundreds of common genetic variants associated with the risk of multifactorial diseases. However, their impact on discrimination and risk prediction is limited. It has been suggested that the identification of gene-gene (G-G) and gene-environment (G-E) interactions would improve disease prediction and facilitate prevention. We conducted a simulation study to explore the potential improvement in discrimination if G-G and G-E interactions exist and are known. We used three diseases (breast cancer, type 2 diabetes, and rheumatoid arthritis) as motivating examples. We show that the inclusion of G-G and G-E interaction effects in risk-prediction models is unlikely to dramatically improve the discrimination ability of these models.

Breast cancer risk prediction and individualised screening based on common genetic variation and breast density measurement

Introduction

Over the last decade several breast cancer risk alleles have been identified which has led to an increased interest in individualised risk prediction for clinical purposes.

Methods

We investigate the performance of an up-to-date 18 breast cancer risk single-nucleotide polymorphisms (SNPs), together with mammographic percentage density (PD), body mass index (BMI) and clinical risk factors in predicting absolute risk of breast cancer, empirically, in a well characterised Swedish case-control study of postmenopausal women. We examined the efficiency of various prediction models at a population level for individualised screening by extending a recently proposed analytical approach for estimating number of cases captured.

Results

The performance of a risk prediction model based on an initial set of seven breast cancer risk SNPs is improved by additionally including eleven more recently established breast cancer risk SNPs (P = 4.69 × 10^-4). Adding mammographic PD, BMI and all 18 SNPs to a Swedish Gail model improved the discriminatory accuracy (the AUC statistic) from 55% to 62%. The net reclassification improvement was used to assess improvement in classification of women into low, intermediate, and high categories of 5-year risk (P = 8.93 × 10^-9). For scenarios we considered, we estimated that an individualised screening strategy based on risk models incorporating clinical risk factors, mammographic density and SNPs, captures 10% more cases than a screening strategy using the same resources, based on age alone. Estimates of numbers of cases captured by screening stratified by age provide insight into how individualised screening programs might appear in practice.

Conclusions

Taken together, genetic risk factors and mammographic density offer moderate improvements to clinical risk factor models for predicting breast cancer.

Translating biomarkers to clinical practice

Biomarkers are the measurable characteristics of an individual that may represent risk factors for a disease or outcome, or that may be indicators of disease progression or of treatment-associated changes. In general, the process by which biomarkers, once identified, might be translated into clinical practice has received scant attention in recent psychiatric literature. A body of work in diagnostic development suggests a framework for evaluating and validating novel biomarkers, but this work may be unfamiliar to clinical and translational researchers in psychiatry. Therefore, this review focuses on the steps that might follow the identification of putative biomarkers. It first addresses standard approaches to characterizing biomarker performance, followed by demonstrations of how a putative biomarker might be shown to have clinical relevance. Finally, it addresses ways in which a biomarker-based test might be validated for clinical application in terms of efficacy and cost-effectiveness.

Annotating individual human genomes

Advances in DNA sequencing technologies have made it possible to rapidly, accurately and affordably sequence entire individual human genomes. As impressive as this ability seems, however, it will not likely amount to much if one cannot extract meaningful information from individual sequence data. Annotating variations within individual genomes and providing information about their biological or phenotypic impact will thus be crucially important in moving individual sequencing projects forward, especially in the context of the clinical use of sequence information. In this paper we consider the various ways in which one might annotate individual sequence variations and point out limitations in the available methods for doing so. It is arguable that, in the foreseeable future, DNA sequencing of individual genomes will become routine for clinical, research, forensic, and personal purposes. We therefore also consider directions and areas for further research in annotating genomic variants.

Risk prediction of complex diseases from family history and known susceptibility loci, with applications for cancer screening

Risk prediction based on genomic profiles has raised a lot of attention recently. However, family history is usually ignored in genetic risk prediction. In this study we proposed a statistical framework for risk prediction given an individual's genotype profile and family history. Genotype information about the relatives can also be incorporated. We allow risk prediction given the current age and follow-up period and consider competing risks of mortality. The framework allows easy extension to any family size and structure. In addition, the predicted risk at any percentile and the risk distribution graphs can be computed analytically. We applied the method to risk prediction for breast and prostate cancers by using known susceptibility loci from genome-wide association studies. For breast cancer, in the population the 10-year risk at age 50 ranged from 1.1% at the 5th percentile to 4.7% at the 95th percentile. If we consider the average 10-year risk at age 50 (2.39%) as the threshold for screening, the screening age ranged from 62 at the 20th percentile to 38 at the 95th percentile (and some never reach the threshold). For women with one affected first-degree relative, the 10-year risks ranged from 2.6% (at the 5th percentile) to 8.1% (at the 95th percentile). For prostate cancer, the corresponding 10-year risks at age 60 varied from 1.8% to 14.9% in the population and from 4.2% to 23.2% in those with an affected first-degree relative. We suggest that for some diseases genetic testing that incorporates family history can stratify people into diverse risk categories and might be useful in targeted prevention and screening.