Value of genetic profiling for the prediction of coronary heart disease

Evaluation of penalized and machine learning methods for asthma disease prediction in the Korean Genome and Epidemiology Study (KoGES)

BACKGROUND

Genome-wide association studies have successfully identified genetic variants associated with human disease. Various statistical approaches based on penalized and machine learning methods have recently been proposed for disease prediction. In this study, we evaluated the performance of several such methods for predicting asthma using the Korean Chip (KORV1.1) from the Korean Genome and Epidemiology Study (KoGES).

RESULTS

First, single-nucleotide polymorphisms were selected via single-variant tests using logistic regression with the adjustment of several epidemiological factors. Next, we evaluated the following methods for disease prediction: ridge, least absolute shrinkage and selection operator, elastic net, smoothly clipped absolute deviation, support vector machine, random forest, boosting, bagging, naïve Bayes, and k-nearest neighbor. Finally, we compared their predictive performance based on the area under the curve of the receiver operating characteristic curves, precision, recall, F1-score, Cohen's Kappa, balanced accuracy, error rate, Matthews correlation coefficient, and area under the precision-recall curve. Additionally, three oversampling algorithms are used to deal with imbalance problems.

CONCLUSIONS

Our results show that penalized methods exhibit better predictive performance for asthma than that achieved via machine learning methods. On the other hand, in the oversampling study, randomforest and boosting methods overall showed better prediction performance than penalized methods.

Methods to Assess Genetic Risk Prediction

It is recognized that traditional risk factors do not identify everyone who will develop cardiovascular disease. There is a growing interest in the discovery of novel biomarkers that will augment the predictive potential of traditional cardiovascular risk factors. The era of genome-wide association studies (GWAS) has resulted in the discovery of common genetic polymorphisms associated with a multitude of cardiovascular traits and raises the possibility that these variants can be used in clinical risk prediction. Assessing and evaluating the new genetic risk markers and quantification of the improvement in risk prediction models that incorporate this information is a major challenge. In this paper we discuss the key metrics that are used to assess prediction models-discrimination, calibration, reclassification, and demonstration on how to calculate and interpret these metrics.

Prediction of Quantitative Traits Using Common Genetic Variants: Application to Body Mass Index

With the success of the genome-wide association studies (GWASs), many candidate loci for complex human diseases have been reported in the GWAS catalog. Recently, many disease prediction models based on penalized regression or statistical learning methods were proposed using candidate causal variants from significant single-nucleotide polymorphisms of GWASs. However, there have been only a few systematic studies comparing existing methods. In this study, we first constructed risk prediction models, such as stepwise linear regression (SLR), least absolute shrinkage and selection operator (LASSO), and Elastic-Net (EN), using a GWAS chip and GWAS catalog. We then compared the prediction accuracy by calculating the mean square error (MSE) value on data from the Korea Association Resource (KARE) with body mass index. Our results show that SLR provides a smaller MSE value than the other methods, while the numbers of selected variables in each model were similar.

Risk Prediction Using Genome-Wide Association Studies on Type 2 Diabetes

The success of genome-wide association studies (GWASs) has enabled us to improve risk assessment and provide novel genetic variants for diagnosis, prevention, and treatment. However, most variants discovered by GWASs have been reported to have very small effect sizes on complex human diseases, which has been a big hurdle in building risk prediction models. Recently, many statistical approaches based on penalized regression have been developed to solve the “large p and small n” problem. In this report, we evaluated the performance of several statistical methods for predicting a binary trait: stepwise logistic regression (SLR), least absolute shrinkage and selection operator (LASSO), and Elastic-Net (EN). We first built a prediction model by combining variable selection and prediction methods for type 2 diabetes using Affymetrix Genome-Wide Human SNP Array 5.0 from the Korean Association Resource project. We assessed the risk prediction performance using area under the receiver operating characteristic curve (AUC) for the internal and external validation datasets. In the internal validation, SLR-LASSO and SLR-EN tended to yield more accurate predictions than other combinations. During the external validation, the SLR-SLR and SLR-EN combinations achieved the highest AUC of 0.726. We propose these combinations as a potentially powerful risk prediction model for type 2 diabetes.

Genetics of coronary artery disease and myocardial infarction

Atherosclerotic coronary artery disease (CAD) comprises a broad spectrum of clinical entities that include asymptomatic subclinical atherosclerosis and its clinical complications, such as angina pectoris, myocardial infarction (MI) and sudden cardiac death. CAD continues to be the leading cause of death in industrialized society. The long-recognized familial clustering of CAD suggests that genetics plays a central role in its development, with the heritability of CAD and MI estimated at approximately 50% to 60%. Understanding the genetic architecture of CAD and MI has proven to be difficult and costly due to the heterogeneity of clinical CAD and the underlying multi-decade complex pathophysiological processes that involve both genetic and environmental interactions. This review describes the clinical heterogeneity of CAD and MI to clarify the disease spectrum in genetic studies, provides a brief overview of the historical understanding and estimation of the heritability of CAD and MI, recounts major gene discoveries of potential causal mutations in familial CAD and MI, summarizes CAD and MI-associated genetic variants identified using candidate gene approaches and genome-wide association studies (GWAS), and summarizes the current status of the construction and validations of genetic risk scores for lifetime risk prediction and guidance for preventive strategies. Potential protective genetic factors against the development of CAD and MI are also discussed. Finally, GWAS have identified multiple genetic factors associated with an increased risk of in-stent restenosis following stent placement for obstructive CAD. This review will also address genetic factors associated with in-stent restenosis, which may ultimately guide clinical decision-making regarding revascularization strategies for patients with CAD and MI.

Genetic diagnosis and prognosis of Alzheimer's disease: challenges and opportunities

Alzheimer's disease (AD), the most common form of dementia in western societies, is a pathologically and clinically heterogeneous disease with a strong genetic component. The recent advances in high-throughput genome technologies allowing for the rapid analysis of millions of polymorphisms in thousands of subjects has significantly advanced our understanding of the genomic underpinnings of AD susceptibility. During the last 5 years, genome-wide association and whole-exome- and whole-genome sequencing studies have mapped more than 20 disease-associated loci, providing insights into the molecular pathways involved in AD pathogenesis and hinting at potential novel therapeutic targets. This review article summarizes the challenges and opportunities of when using genomic information for the diagnosis and prognosis of AD.

Expanding Horizons for Abdominal Aortic Aneurysms

Recent technological advances have allowed researchers to interrogate the genetic basis of abdominal aortic aneurysms in great detail. The results from these studies are expected to transform our understanding of this complex disease with both multiple genetic and environmental risk factors. Clinicians need to keep abreast of these genetic findings and understand the implications for their practice. Patients will become increasingly informed on genetic risk, and a new era of individualized risk assessment for AAA is just beginning. This brief update aims to provide the clinician with a succinct précis of the recent progress in this area.

Age-related macular degeneration treatment in the era of molecular medicine

Age-related macular degeneration (AMD) is the leading cause of irreversible blindness in the developed world. The quality of life of both patients and families is impacted by this prevalent disease. Previously, macular degeneration had no known effective treatment. Today, vitamins for non-exudative AMD and intravitreal injection of medications for its exudative form are primary forms of current treatment. Modern advances in molecular science give rise to new possibilities of disease management. In the year 2003 the sequencing of the entire human genome was completed. Since that time, genes such as complement factor H, high-temperature requirement factor A1, and age-relateed maculopathy susceptibility 2 have been discovered and associated with a higher risk of AMD. A patient’s genetic make-up may dictate the effectiveness of current or future therapeutic options. In addition, utilizing genetic data and incorporating it into new treatments (such as viral vectors) may lead to longer-lasting (or permanent) VEGF blockade and specific targeting of complement related genes. There have also been considerable advances in stem cell directed treatment of AMD. Retinal pigment epithelial (RPE) cells can be derived from human embryonic stem cells, induced pluripotent stem cells, or adult human RPE stem cells. Utilizing animal models of RPE and retinal degeneration, stem cell-derived RPE cells have been successfully implanted into the subretinal space. They have been injected as a cell mass or as a pre-prepared monolayer on a thin membrane. Visual recovery has been demonstrated in a retinal dystrophic rat model. Preliminary data on 2 human subjects also demonstrates possible early visual benefit from transplantation of stem cell-derived RPE. As more data is published, and as differentiation and implantation techniques are optimized, the stabilization and possible improvement of vision in individuals with non-exudative macular becomes a real possibility. We conclude that the technologic advances that continue to unfold in both genetic and stem cell research offer optimism in the future treatment of AMD.

Genetic risk score of NOS gene variants associated with myocardial infarction correlates with coronary incidence across Europe

Coronary artery disease (CAD) mortality and morbidity is present in the European continent in a four-fold gradient across populations, from the South (Spain and France) with the lowest CAD mortality, towards the North (Finland and UK). This observed gradient has not been fully explained by classical or single genetic risk factors, resulting in some cases in the so called Southern European or Mediterranean paradox. Here we approached population genetic risk estimates using genetic risk scores (GRS) constructed with single nucleotide polymorphisms (SNP) from nitric oxide synthases (NOS) genes. These SNPs appeared to be associated with myocardial infarction (MI) in 2165 cases and 2153 controls. The GRSs were computed in 34 general European populations. Although the contribution of these GRS was lower than 1% between cases and controls, the mean GRS per population was positively correlated with coronary incidence explaining 65–85% of the variation among populations (67% in women and 86% in men). This large contribution to CAD incidence variation among populations might be a result of colinearity with several other common genetic and environmental factors. These results are not consistent with the cardiovascular Mediterranean paradox for genetics and support a CAD genetic architecture mainly based on combinations of common genetic polymorphisms. Population genetic risk scores is a promising approach in public health interventions to develop lifestyle programs and prevent intermediate risk factors in certain subpopulations with specific genetic predisposition.

The ENCODE project and perspectives on pathways

The recently completed ENCODE project is a new source of information on metabolic activity, unveiling knowledge about evolution and similarities among species, refuting the myth that most DNA is "junk" and has no actual function. With this expansive resource comes a challenge: integrating these new layers of information into our current knowledge of single-nucleotide polymorphisms and previously described metabolic pathways with the aim of discovering new genes and pathways related to human diseases and traits. Further, we must determine which computational methods will be most useful in this pursuit. In this paper, we speculate over the possible methods that will emerge in this new, challenging field.

Molecular genetic overlap in bipolar disorder, schizophrenia, and major depressive disorder

OBJECTIVES

Genome-wide association studies (GWAS) in complex phenotypes, including psychiatric disorders, have yielded many replicated findings, yet individual markers account for only a small fraction of the inherited differences in risk. We tested the performance of polygenic models in discriminating between cases and healthy controls and among cases with distinct psychiatric diagnoses.

METHODS

GWAS results in bipolar disorder (BD), major depressive disorder (MDD), schizophrenia (SZ), and Parkinson's disease (PD) were used to assign weights to individual alleles, based on odds ratios. These weights were used to calculate allele scores for individual cases and controls in independent samples, summing across many single nucleotide polymorphisms (SNPs). How well allele scores discriminated between cases and controls and between cases with different disorders was tested by logistic regression.

RESULTS

Large sets of SNPs were needed to achieve even modest discrimination between cases and controls. The most informative SNPs were overlapping in BD, SZ, and MDD, with correlated effect sizes. Little or no overlap was seen between allele scores for psychiatric disorders and those for PD.

CONCLUSIONS

BD, SZ, and MDD all share a similar polygenic component, but the polygenic models tested lack discriminative accuracy and are unlikely to be useful for clinical diagnosis.

Bringing genome-wide association findings into clinical use

Genome-wide association studies (GWASs) have been heralded as a major advance in biomedical discovery, having identified ~2,000 robust associations with complex diseases since 2005. Despite this success, they have met considerable scepticism regarding their clinical applicability; this scepticism arises from such aspects as the modest effect sizes of associated variants and their unclear functional consequences. There are, however, promising examples of GWAS findings that will or that may soon be translated into clinical care. These examples include variants identified through GWASs that provide strongly predictive or prognostic information or that have important pharmacological implications; these examples may illustrate promising approaches to wider clinical application.

Predictive accuracy of risk factors and markers: a simulation study of the effect of novel markers on different performance measures for logistic regression models

The change in c-statistic is frequently used to summarize the change in predictive accuracy when a novel risk factor is added to an existing logistic regression model. We explored the relationship between the absolute change in the c-statistic, Brier score, generalized R(2) , and the discrimination slope when a risk factor was added to an existing model in an extensive set of Monte Carlo simulations. The increase in model accuracy due to the inclusion of a novel marker was proportional to both the prevalence of the marker and to the odds ratio relating the marker to the outcome but inversely proportional to the accuracy of the logistic regression model with the marker omitted. We observed greater improvements in model accuracy when the novel risk factor or marker was uncorrelated with the existing predictor variable compared with when the risk factor has a positive correlation with the existing predictor variable. We illustrated these findings by using a study on mortality prediction in patients hospitalized with heart failure. In conclusion, the increase in predictive accuracy by adding a marker should be considered in the context of the accuracy of the initial model.

Phenotype prediction from genome-wide association studies: application to smoking behaviors

Background

A great success of the genome wide association study enabled us to give more attention on the personal genome and clinical application such as diagnosis and disease risk prediction. However, previous prediction studies using known disease associated loci have not been successful (Area Under Curve 0.55 ~ 0.68 for type 2 diabetes and coronary heart disease). There are several reasons for poor predictability such as small number of known disease-associated loci, simple analysis not considering complexity in phenotype, and a limited number of features used for prediction.

Methods

In this research, we investigated the effect of feature selection and prediction algorithm on the performance of prediction method thoroughly. In particular, we considered the following feature selection and prediction methods: regression analysis, regularized regression analysis, linear discriminant analysis, non-linear support vector machine, and random forest. For these methods, we studied the effects of feature selection and the number of features on prediction. Our investigation was based on the analysis of 8,842 Korean individuals genotyped by Affymetrix SNP array 5.0, for predicting smoking behaviors.

Results

To observe the effect of feature selection methods on prediction performance, selected features were used for prediction and area under the curve score was measured. For feature selection, the performances of support vector machine (SVM) and elastic-net (EN) showed better results than those of linear discriminant analysis (LDA), random forest (RF) and simple logistic regression (LR) methods. For prediction, SVM showed the best performance based on area under the curve score. With less than 100 SNPs, EN was the best prediction method while SVM was the best if over 400 SNPs were used for the prediction.

Conclusions

Based on combination of feature selection and prediction methods, SVM showed the best performance in feature selection and prediction.

Do genome-wide association scans have potential for translation?

The success of genome-wide association studies (GWAS) in identifying replicating associations has greatly contributed to understanding of the genetic aetiology of complex diseases. This review discusses and provides examples of the potential of GWAS findings to be translated into clinical practice, i.e., diagnosis, prediction, prognosis, novel treatments and response to treatment of common diseases. The biological insights afforded by newly-identified robust associations represent the largest, albeit indirect, translational contribution of GWAS.

Racial/ethnic variation in the association of lipid-related genetic variants with blood lipids in the US adult population

BACKGROUND

Genome-wide association studies (GWAS) have identified a number of single-nucleotide polymorphisms (SNPs) associated with serum lipid level in populations of European descent. The individual and the cumulative effect of these SNPs on blood lipids are largely unclear for the US population.

METHODS AND RESULTS

Using data from the second phase (1991-1994) of the Third National Health and Nutrition Examination Survey (NHANES III), a nationally representative survey of the US population, we examined associations of 57 GWAS-identified or well-established lipid-related genetic loci with plasma concentrations of high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol, total cholesterol, triglycerides, total cholesterol/HDL-C ratio, and non-HDL-C. We used multivariable linear regression to examine single SNP associations and the cumulative effect of multiple SNPs (using a genetic risk score [GRS]) on blood lipid levels. Analyses were conducted in adults from each of the 3 major racial/ethnic groups in the United States: non-Hispanic whites (n=2296), non-Hispanic blacks (n=1699), and Mexican Americans (n=1713). Allele frequencies for all SNPs varied significantly by race/ethnicity, except rs3764261 in CETP. Individual SNPs had very small effects on lipid levels, effects that were generally consistent in direction across racial/ethnic groups. More GWAS-validated SNPs were replicated in non-Hispanic whites (<67%) than in non-Hispanic blacks (<44%) or Mexican Americans (<44%). GRSs were strongly associated with increased lipid levels in each racial/ethnic group. The combination of all SNPs into a weighted GRS explained no more than 11% of the total variance in blood lipid levels.

CONCLUSIONS

Our findings show that the combined association of SNPs, based on a GRS, was strongly associated with increased blood lipid measures in all major race/ethnic groups in the United States, which may help in identifying subgroups with a high risk for an unfavorable lipid profile.

Alzheimer's disease genetics: current knowledge and future challenges

Alzheimer's disease (AD) is highly heritable, but genetically complex. Recently, three large-scale genome-wide association studies have made substantial breakthroughs in disentangling the genetic architecture of the disease. These studies combined include data from over 43 000 independent individuals and provide compelling evidence that variants in four novel susceptibility genes (CLU, PICALM, CR1, BIN1) are associated with disease risk. These findings are tremendously exciting, not only in providing new avenues for exploration, but also highlighting the potential for further gene discovery when larger samples are analysed. Here we discuss progress to date in identifying risk genes for dementia, ways forward and how current findings are refining previous ideas and defining new putative primary disease mechanisms.

Practice of personalized primary prevention of lifestyle diseases: associated problems and issues in Japan

This article explores the current situation of personalized primary prevention of lifestyle diseases based on individuals' genome information in Japan. The Japanese healthcare system and regulatory framework of genetic testing have been reviewed because they relate to the issue. From our perspective, clinics and health check-up centers in Japan are appropriate centers for implementing measures for primary prevention of lifestyle diseases. For example, we researched a predictive genetic testing program for diabetic complications and metabolic syndrome. Based on our interviews with physicians and experts, we identified the following factors that are critical to the effective use of genetic testing for primary prevention of lifestyle diseases: an institutional framework for evaluating the credibility of predictive genetic testing; psychological effects (both positive and negative) of predictive genetic testing; and education of physicians and other health professionals in lifestyle disease genetics and effective communication with patients.

Points to consider in assessing and appraising predictive genetic tests

The use of predictive genetic tests is expanding rapidly. Given limited health care budgets and few national coverage decisions specifically for genetic tests, evidence of benefits and harms is a key requirement in decision making; however, assessing the benefits and harms of genetic tests raises a number of challenging issues. Frequently, evidence of medical benefits and harms is limited due to practical and ethical limitations of conducting meaningful clinical trials. Also, clinical endpoints frequently do not capture the benefit appropriately because the main purpose of many genetic tests is personal utility of knowing the test results, and costs of the tests and counseling can be insufficient indicators of the total costs of care. This study provides an overview of points to consider for the assessment of benefits and harms from genetic tests in an ethically and economically reflected manner. We discuss whether genetic tests are sufficiently exceptional to warrant exceptional methods for assessment and appraisal.

Being more realistic about the public health impact of genomic medicine

Wayne Hall and colleagues discuss the limitations of genomic risk prediction for population-level preventive health care.

Genetic variants and their interactions in the prediction of increased pre-clinical carotid atherosclerosis: the cardiovascular risk in young Finns study

The relative contribution of genetic risk factors to the progression of subclinical atherosclerosis is poorly understood. It is likely that multiple variants are implicated in the development of atherosclerosis, but the subtle genotypic and phenotypic differences are beyond the reach of the conventional case-control designs and the statistical significance testing procedures being used in most association studies. Our objective here was to investigate whether an alternative approach--in which common disorders are treated as quantitative phenotypes that are continuously distributed over a population--can reveal predictive insights into the early atherosclerosis, as assessed using ultrasound imaging-based quantitative measurement of carotid artery intima-media thickness (IMT). Using our population-based follow-up study of atherosclerosis precursors as a basis for sampling subjects with gradually increasing IMT levels, we searched for such subsets of genetic variants and their interactions that are the most predictive of the various risk classes, rather than using exclusively those variants meeting a stringent level of statistical significance. The area under the receiver operating characteristic curve (AUC) was used to evaluate the predictive value of the variants, and cross-validation was used to assess how well the predictive models will generalize to other subsets of subjects. By means of our predictive modeling framework with machine learning-based SNP selection, we could improve the prediction of the extreme classes of atherosclerosis risk and progression over a 6-year period (average AUC 0.844 and 0.761), compared to that of using conventional cardiovascular risk factors alone (average AUC 0.741 and 0.629), or when combined with the statistically significant variants (average AUC 0.762 and 0.651). The predictive accuracy remained relatively high in an independent validation set of subjects (average decrease of 0.043). These results demonstrate that the modeling framework can utilize the "gray zone" of genetic variation in the classification of subjects with different degrees of risk of developing atherosclerosis.

The genetics of cardiovascular disease: new insights from emerging approaches

The prospect that sequencing the human genome would see rapid translation of a greater understanding of cardiovascular genetics into novel diagnostics and therapeutics has so far met with only limited success. However, diverse technological advances and exploitation of novel animal models of cardiovascular development and disease are providing ever more insight into cardiovascular diseases and development, and bring closer the prospect of 'post-genomic' diagnostics and therapies. Here we review some of these emerging approaches (genome wide association studies, deep sequencing, microRNA regulation, and zebrafish as a model of cardiovascular disease and development) and discuss their potential for finally fulfilling the promise of application to clinical cardiovascular medicine.

The pursuit of susceptibility genes for Alzheimer's disease: progress and prospects

The recent discoveries in genome-wide association studies (GWAS) of novel susceptibility loci (CLU, CR1 and PICALM) for Alzheimer's disease (AD) have elicited considerable interest in the AD community. But what are the implications of these purely epidemiological findings for our understanding of disease etiology and patient care? In this review, we attempt to place these findings in the context of current and future AD genetics research. CLU, CR1 and PICALM support existing hypotheses about the amyloid, lipid, chaperone and chronic inflammatory pathways in AD pathogenesis. We discuss how these and future findings can be translated into efforts to ameliorate patient care by genetic profiling for risk prediction and pharmacogenetics and by guiding drug development.

From disease association to risk assessment: an optimistic view from genome-wide association studies on type 1 diabetes

Genome-wide association studies (GWAS) have been fruitful in identifying disease susceptibility loci for common and complex diseases. A remaining question is whether we can quantify individual disease risk based on genotype data, in order to facilitate personalized prevention and treatment for complex diseases. Previous studies have typically failed to achieve satisfactory performance, primarily due to the use of only a limited number of confirmed susceptibility loci. Here we propose that sophisticated machine-learning approaches with a large ensemble of markers may improve the performance of disease risk assessment. We applied a Support Vector Machine (SVM) algorithm on a GWAS dataset generated on the Affymetrix genotyping platform for type 1 diabetes (T1D) and optimized a risk assessment model with hundreds of markers. We subsequently tested this model on an independent Illumina-genotyped dataset with imputed genotypes (1,008 cases and 1,000 controls), as well as a separate Affymetrix-genotyped dataset (1,529 cases and 1,458 controls), resulting in area under ROC curve (AUC) of approximately 0.84 in both datasets. In contrast, poor performance was achieved when limited to dozens of known susceptibility loci in the SVM model or logistic regression model. Our study suggests that improved disease risk assessment can be achieved by using algorithms that take into account interactions between a large ensemble of markers. We are optimistic that genotype-based disease risk assessment may be feasible for diseases where a notable proportion of the risk has already been captured by SNP arrays.