A testing framework for identifying susceptibility genes in the presence of epistasis

Functional polymorphism of detoxification gene NQO1 predicts intensity of empirical treatment of childhood asthma

The management of moderate to severe childhood asthma remains empirical. Genotypic variation has been proposed as a way to tailor specific pharmaceutical regimens to individual patients. The objective of this study was to determine the factors associated with asthma treatment progression, including functional polymorphisms of phase II detoxification enzymes, demographics, and environmental factors. In a study of 120 asthmatic children cared for in a single pediatric pulmonary practice, intensity of medical treatment over the year prior was modeled as a function of null mutations of glutathione S transferase (GST) M1 and T1, ile105val variant of GSTP1, and pro187ser variant of NAD(P)H:quinone oxidoreductase 1 (NQO1). The model included demographics, medical information, and environmental factors obtained via questionnaire analyzed with multivariate logistic regression and artificial neural networks. Multivariate logistic regression with bootstrapped validation identified a polymorphic variant of NQO1 as significantly contributing to increasing the odds of receiving more aggressive medical therapy (odds ratio, 11.56; p=0.0001). Parent income and education inversely correlated with medical treatment (odds ratio, 1.50; p=0.001 and odds ratio, 0.375; p=0.002, respectively). Age and reporting restricted physical activity due to asthma also impacted medical treatment (odds ratio, 0.63; p=0.0001 and odds ratio, 5.90; p=0.004, respectively). The optimism-adjusted discriminative ability (c-index) of the model was 0.881 (close to Bayes optimum of 0.902) with 80% overall classification accuracy. Our study supports the role of NQO1 polymorphism as an important factor determining the intensity of medical therapy in asthmatic children after adjusting for significance relating to parental income and education level, age, and restricted physical activity. Asthmatic children with a functional polymorphism of NQO1 may require more intensive pharmaceutical treatment to effectively control their asthma.

Association between ADORA2A and DRD2 polymorphisms and caffeine-induced anxiety

Caffeine produces mild psychostimulant and sometimes anxiogenic effects by antagonizing adenosine at A(1) and A(2A) receptors, and perhaps through interactions with other transmitter systems. Adenosine receptors are colocalized and functionally interact with dopamine receptors in the brain. Thus, functional polymorphisms in the genes for either adenosine or dopamine receptors may affect responses to caffeine. In this study, we examined associations between self-reported anxiogenic effects of caffeine and variation in the genes for A(2A) (ADORA2A) and DRD(2) (DRD2) receptors. Healthy male and female individuals (n=102), who consumed less than 300 mg caffeine per week, ingested capsules containing 0, 50, 150, and 450 mg caffeine under double-blind conditions in four separate experimental sessions. Subjective anxiety was measured before and at repeated times after capsules were consumed. At the 150 mg dose of caffeine, we found a significant association between caffeine-induced anxiety (Visual Analog Scales, VAS) and ADORA2A rs5751876 (1976C/T), rs2298383 (intron 1a) and rs4822492 (3'-flank), and DRD2 rs1110976 (intron 6). Caffeine-induced anxiety (VAS) was also associated with two-loci interactions of selected ADORA2A and DRD2 polymorphisms. The lowest dose of caffeine did not increase ratings of anxiety while the highest dose increased anxiety in the majority of subjects. These findings provide support for an association between an ADORA2A polymorphism and self-reported anxiety after a moderate dose of caffeine. It is likely that other ADORA2A and DRD2 polymorphisms also contribute to responses to caffeine.

Ozone exposure, antioxidant genes, and lung function in an elderly cohort: VA normative aging study

BACKGROUND

Ozone (O3) exposure is known to cause oxidative stress. This study investigated the acute effects of O(3) on lung function in the elderly, a suspected risk group. It then investigated whether genetic polymorphisms of antioxidant genes (heme oxygenase-1 (HMOX1) and glutathione S-transferase pi (GSTP1)) modified these associations.

METHODS

1100 elderly men from the Normative Aging Study were examined whose lung function (forced vital capacity (FVC) and forced expiratory volume in 1 second (FEV1)) was measured every 3 years from 1995 to 2005. The study genotyped the GSTP1 Ile105Val and Ala114Val polymorphisms and the (GT)n repeat polymorphism in the HMOX1 promoter, classifying repeats as short (n<25) or long (n> or =25). Ambient O(3) was measured continuously at locations in the Greater Boston area. Mixed linear models were used, adjusting for known confounders.

RESULTS

A 15 ppb increase in O(3) during the previous 48 h was associated with a 1.25% decrease in FEV(1) (95% CI: -1.96% to -0.54%). This estimated effect was worsened with either the presence of a long (GT)n repeat in HMOX1 (-1.38%, 95% CI: -2.11% to -0.65%) or the presence of an allele coding for Val105 in GSTP1 (-1.69%, 95% CI: -2.63% to -0.75%). A stronger estimated effect of O(3) on FEV(1) was found in subjects carrying both the GSTP1 105Val variant and the HMOX1 long (GT)n repeat (-1.94%, 95% CI: -2.89% to -0.98%). Similar associations were also found between FVC and O(3) exposure.

CONCLUSIONS

Our results suggest that O(3) has an acute effect on lung function in the elderly, and the effects may be modified by the presence of specific polymorphisms in antioxidant genes.

Detecting pathway-based gene-gene and gene-environment interactions in pancreatic cancer

Data mining and data reduction methods to detect interactions in epidemiologic data are being developed and tested. In these analyses, multifactor dimensionality reduction, focused interaction testing framework, and traditional logistic regression models were used to identify potential interactions with up to three factors. These techniques were used in a population-based case-control study of pancreatic cancer from the San Francisco Bay Area (308 cases, 964 controls). From 7 biochemical pathways, along with tobacco smoking, 26 polymorphisms in 20 genes were included in these analyses. Combinations of genetic markers and cigarette smoking were identified as potential risk factors for pancreatic cancer, including genes in base excision repair (OGG1), nucleotide excision repair (XPD, XPA, XPC), and double-strand break repair (XRCC3). XPD.751, XPD.312, and cigarette smoking were the best single-factor predictors of pancreatic cancer risk, whereas XRCC3.241*smoking and OGG1.326*XPC.PAT were the best two-factor predictors. There was some evidence for a three-factor combination of OGG1.326*XPD.751*smoking, but the covariate-adjusted relative-risk estimates lacked precision. Multifactor dimensionality reduction and focused interaction testing framework showed little concordance, whereas logistic regression allowed for covariate adjustment and model confirmation. Our data suggest that multiple common alleles from DNA repair pathways in combination with cigarette smoking may increase the risk for pancreatic cancer, and that multiple approaches to data screening and analysis are necessary to identify potentially new risk factor combinations.

Clock genes may influence bipolar disorder susceptibility and dysfunctional circadian rhythm

Several previous studies suggest that dysfunction of circadian rhythms may increase susceptibility to bipolar disorder (BP). We conducted an association study of five circadian genes (CRY2, PER1-3, and TIMELESS) in a family collection of 36 trios and 79 quads (Sample I), and 10 circadian genes (ARNTL, ARNTL2, BHLHB2, BHLHB3, CLOCK, CRY1, CSNK1D, CSNK1E, DBP, and NR1D1) in an extended family collection of 70 trios and 237 quads (Sample II), which includes the same 114 families but not necessarily the same individuals as Sample I. In Sample II, the Sibling-Transmission Disequilibrium Test (sib-tdt) analysis showed nominally significant association of BP with three SNPs within or near the CLOCK gene (rs534654, P = 0.0097; rs6850524, P = 0.012; rs4340844, P = 0.015). In addition, SNPs in the ARNTL2, CLOCK, DBP, and TIMELESS genes and haplotypes in the ARNTL, CLOCK, CSNK1E, and TIMELESS genes showed suggestive evidence of association with several circadian phenotypes identified in BP patients. However, none of these associations reached gene-wide or experiment-wide significance after correction for multiple-testing. A multi-locus interaction between rs6442925 in the 5' upstream of BHLHB2, rs1534891 in CSNK1E, and rs534654 near the 3' end of the CLOCK gene, however, is significantly associated with BP (P = 0.00000172). It remains significant after correcting for multiple testing using the False Discovery Rate method. Our results indicate an interaction between three circadian genes in susceptibility to bipolar disorder.

Neurotransmission and bipolar disorder: a systematic family-based association study

Neurotransmission pathways/systems have been proposed to be involved in the pathophysiology and treatment of bipolar disorder for over 40 years. In order to test the hypothesis that common variants of genes in one or more of five neurotransmission systems confer risk for bipolar disorder, we analyzed 1,005 tag single nucleotide polymorphisms in 90 genes from dopaminergic, serotonergic, noradrenergic, GABAergic, and glutamatergic neurotransmitter systems in 101 trios and 203 quads from Caucasian bipolar families. Our sample has 80% power to detect ORs >or= 1.82 and >or=1.57 for minor allele frequencies of 0.1 and 0.5, respectively. Nominally significant allelic and haplotypic associations were found for genes from each neurotransmission system, with several reaching gene-wide significance (allelic: GRIA1, GRIN2D, and QDPR; haplotypic: GRIN2C, QDPR, and SLC6A3). However, none of these associations survived correction for multiple testing in an individual system, or in all systems considered together. Significant single nucleotide polymorphism associations were not found with sub-phenotypes (alcoholism, psychosis, substance abuse, and suicide attempts) or significant gene-gene interactions. These results suggest that, within the detectable odds ratios of this study, common variants of the selected genes in the five neurotransmission systems do not play major roles in influencing the risk for bipolar disorder or comorbid sub-phenotypes.

Heritability of self-reported asthma and allergy: a study in adult Dutch twins, siblings and parents

The present study assessed the prevalence of asthma and allergy, and estimated the importance of genetic and environmental influences on asthma and allergy liability and their association. Longitudinal data on self-reported, doctor-diagnosed asthma and allergy were collected in over 14,000 individuals registered with the Netherlands Twin Register. Structural equation modeling was used for univariate and bivariate genetic analyses on data from twins, their siblings, and parents. Results showed no sex, age, and minimal birth cohort effects for asthma prevalence (11.8%). For allergy, prevalence was higher in women (19.8%) than in men (13.9%). Allergy prevalence at ages 22, 23, and 24 years increased from the 1970 to the 1980 birth cohort. The prevalence of allergy, but not of asthma, was higher in nontwin siblings than in twins. No assortative mating was observed. High (broad-sense) heritabilities were found for asthma (75%) and allergy (66%), with evidence for nonadditive genetic effects in asthma. The association between asthma and allergy (correlation=.65) was largely due to common genes (70%). No sex differences in genetic architecture were found. In conclusion, the prevalence of allergy but not of asthma increased in recent years. Individual differences in the liability to asthma, allergy and their co-occurrence are for a large part accounted for by differences in genetic background. Nonadditive gene action is important, which may have consequences for gene hunting strategies.

A novel strategy for detecting multiple loci in Genome-Wide Association Studies of complex diseases

Large-scale Genome-Wide Association Studies (GWAS) for complex diseases are increasingly common, due to recent advances in genotyping technology. Gene-gene interactions play an important role in the etiology of complex diseases and have to be addressed in GWAS. In this paper, an efficient strategy based on two-stage analysis is proposed. It combines a single-locus approach with a Goodness-Of-Fit (GOF) test in stage one, and selects a promising subset of SNPs to be modelled using a full interaction model in stage two. Extensive simulations using different disease models with different levels of epistasis demonstrate that it achieves higher power than existing approaches.

Ability of epistatic interactions of cytokine single-nucleotide polymorphisms to predict susceptibility to disease subsets in systemic sclerosis patients

OBJECTIVE

Gene-gene interaction, or epistasis, is considered a ubiquitous component of complex human diseases such as systemic sclerosis (SSc). Epistasis is difficult to model by traditional parametric approaches; therefore, nonparametric computational algorithms, such as multifactor dimensionality reduction (MDR), have been developed.

METHODS

A total of 242 consecutive unrelated Italian SSc patients and an equal number of well-matched healthy controls were genotyped for 22 cytokine single-nucleotide polymorphisms (SNPs; 13 cytokine genes). The distribution of the SNPs between controls and SSc patients, controls and limited cutaneous SSc (lcSSc) patients, and controls and diffuse cutaneous SSc (dcSSc) patients was tested by the MDR constructive induction algorithm and by focused interaction testing framework (FITF), a logistic regression-based approach.

RESULTS

None of the studied SNPs had main independent effects on SSc or disease subset susceptibility, therefore no epistatic interaction was detectable by FITF. The MDR analysis showed a significant epistatic interaction among the interleukin-2 (IL-2) G-330T, IL-6 C-174G, and interferon-gamma AUTR5644T SNPs and the IL-1 receptor Cpst1970T, IL-6 Ant565G, and IL-10 C-819T SNPs in lcSSc and dcSSc susceptibility, respectively. The relevance of the single multilocus attributes constructed by the MDR inductive algorithm was then confirmed by the parametric approach (P < 0.001 for both controls versus lcSSc patients and controls versus dcSSc patients).

CONCLUSION

We provide evidence for gene-gene interaction among cytokine SNPs in the context of SSc. The interaction among cytokine SNPs with a profibrotic or a regulatory function on profibrotic interleukins is relevant to the susceptibility to SSc subsets and it appears to be more important than the contribution of any single cytokine SNP.

Comparison of approaches for machine-learning optimization of neural networks for detecting gene-gene interactions in genetic epidemiology

The detection of genotypes that predict common, complex disease is a challenge for human geneticists. The phenomenon of epistasis, or gene-gene interactions, is particularly problematic for traditional statistical techniques. Additionally, the explosion of genetic information makes exhaustive searches of multilocus combinations computationally infeasible. To address these challenges, neural networks (NN), a pattern recognition method, have been used. One limitation of the NN approach is that its success is dependent on the architecture of the network. To solve this, machine-learning approaches have been suggested to evolve the best NN architecture for a particular data set. In this study we provide a detailed technical description of the use of grammatical evolution to optimize neural networks (GENN) for use in genetic association studies. We compare the performance of GENN to that of a previous machine-learning NN application--genetic programming neural networks in both simulated and real data. We show that GENN greatly outperforms genetic programming neural networks in data sets with a large number of single nucleotide polymorphisms. Additionally, we demonstrate that GENN has high power to detect disease-risk loci in a range of high-order epistatic models. Finally, we demonstrate the scalability of the GENN method with increasing numbers of variables--as many as 500,000 single nucleotide polymorphisms.

An ensemble learning approach jointly modeling main and interaction effects in genetic association studies

Complex diseases are presumed to be the results of interactions of several genes and environmental factors, with each gene only having a small effect on the disease. Thus, the methods that can account for gene-gene interactions to search for a set of marker loci in different genes or across genome and to analyze these loci jointly are critical. In this article, we propose an ensemble learning approach (ELA) to detect a set of loci whose main and interaction effects jointly have a significant association with the trait. In the ELA, we first search for "base learners" and then combine the effects of the base learners by a linear model. Each base learner represents a main effect or an interaction effect. The result of the ELA is easy to interpret. When the ELA is applied to analyze a data set, we can get a final model, an overall P-value of the association test between the set of loci involved in the final model and the trait, and an importance measure for each base learner and each marker involved in the final model. The final model is a linear combination of some base learners. We know which base learner represents a main effect and which one represents an interaction effect. The importance measure of each base learner or marker can tell us the relative importance of the base learner or marker in the final model. We used intensive simulation studies as well as a real data set to evaluate the performance of the ELA. Our simulation studies demonstrated that the ELA is more powerful than the single-marker test in all the simulation scenarios. The ELA also outperformed the other three existing multi-locus methods in almost all cases. In an application to a large-scale case-control study for Type 2 diabetes, the ELA identified 11 single nucleotide polymorphisms that have a significant multi-locus effect (P-value=0.01), while none of the single nucleotide polymorphisms showed significant marginal effects and none of the two-locus combinations showed significant two-locus interaction effects.

Interaction of gender, hypertension, and the angiotensinogen gene haplotypes on the risk of coronary artery disease in a large angiographic cohort

There is increasing evidence suggesting the importance of evaluating gene-environment interactions in the genetic study of coronary artery disease (CAD). We investigated the association of multiple single nucleotide polymorphisms in the angiotensinogen (AGT) gene with CAD, considering the interaction between the genetic and non-genetic factors, using a larger and ethnically homogeneous angiographic cohort. A total of 1254 consecutive patients who underwent cardiac catheterization (735 with CAD and 519 without) were recruited. T174M (rs4762), M235T (rs699), G-6A, A-20C, G-152A, and G-217A polymorphisms of the AGT gene were genotyped. We used a regression approach based on a generalized linear model to evaluate haplotype effects defined by the multilocus data and detection of gene-environment interaction by incorporating interaction terms in the model. We found significant differences in global AGT gene haplotype profile between patients with and without CAD (the global score statistic=25.411, P=0.008). Significant interactions between AGT gene haplotypes, gender and hypertension were detected. We also used haplotype counting to directly estimate the odds ratio of each AGT gene haplotype, and found that the effects of haplotypes were markedly different in subgroups defined by gender and hypertension, providing strong evidence of gene-environment interaction. Female gender synergistically enhances (or male gender reverses) the effects of AGT gene haplotypes on the risk of CAD in the presence of hypertension. In conclusion, the effect of AGT gene haplotypes on the risk of CAD was significantly increased in women with hypertension, which highlights the importance of evaluating gene-environment interactions in the genetic study of CAD.

Gene expression profiling of in vitro cultured macrophages after exposure to the respiratory sensitizer hexamethylene diisocyanate

Occupational exposure to chemicals is one of the main causes of respiratory allergy and asthma. Identification of chemicals that trigger allergic asthma is difficult as underlying processes and specific markers have not yet been clearly defined. Moreover, adequate classification of the respiratory toxicity of chemicals is hampered due to the lack of validated in vivo and in vitro test methods. The study of differential gene expression profiles in appropriate human in vitro cell systems is a promising approach to identify selective markers for respiratory allergy. As alveolar macrophages display important immunological and inflammatory properties in response to foreign substances in the lung, we aimed at gaining more insight in changes of human macrophages transcriptome and to identify selective genetic markers for respiratory sensitization in response to hexamethylene diisocyanate (HDI). In vitro cultures of human THP-1 cells were differentiated into macrophages and exposed to 55 microg/ml HDI for 6 and 10h. Using human oligonucleotide microarrays, changes were observed in the expression of genes that are involved in diverse biological and molecular processes, including detoxification, oxidative stress, cytokine signaling, and apoptosis, which can lead to the development of asthma. These genes are possible markers for respiratory sensitization caused by isocyanates.

Composite measure of linkage disequilibrium for testing interaction between unlinked loci

Widely used statistical interaction models essentially treated the interaction effect as a residual term and hence are likely to limit the power to detect interaction. Alternatively, interactions between two loci can be understood as irreducible dependencies between loci causing disease or viewed as the linkage disequilibrium (LD) between them. This motivated the development of LD-based statistics for the detection of interaction between two loci. Although LD-based statistics have demonstrated high power to detect interaction between two loci, in general, linkage phase information of marker loci for unrelated individuals is unknown. To overcome this limitation, we classify the interaction between two loci into intragametic interaction that characterizes interaction of two alleles from different loci on the same haplotype and intergametic interaction that characterizes the interaction of two alleles from different loci on different haplotypes. Then we show that intragametic and intergametic interaction will lead to the corresponding intragametic and intergametic LD. This stimulates the use of composite measure of LD for developing statistics to detect interaction between two unlinked loci. To study the validity of the composite LD-based statistic for testing interaction, we estimate its type 1 error rates by simulation. To evaluate the performance of the composite LD-based statistic for detection of interaction between two loci, we compare its power with logistic regression and apply it to two real examples. The preliminary results demonstrate that the composite LD-based statistic is a strong alternative to the logistic regressions and the intragametic LD-based statistic for the detection of interaction between two unlinked loci.

Systems genetics of alcoholism

Alcoholism is a common disease resulting from the complex interaction of genetic, social, and environmental factors. Interest in the high heritability of alcoholism has resulted in many studies of how single genes, as well as an individual's entire genetic content (i.e., genome) and the proteins expressed by the genome, influence alcoholism risk. The use of large-scale methods to identify and characterize genetic material (i.e., high-throughput technologies) for data gathering and analysis recently has made it possible to investigate the complexity of the genetic architecture of susceptibility to common diseases such as alcoholism on a systems level. Systems genetics is the study of all genetic variations, their interactions with each other (i.e., epistasis), their interactions with the environment (i.e., plastic reaction norms), their relationship with interindividual variation in traits that are influenced by many genes and contribute to disease susceptibility (i.e., intermediate quantitative traits or endophenotypes) defined at different levels of hierarchical biochemical and physiological systems, and their relationship with health and disease. (An endophenotype is a genetically determined trait [i.e., phenotype] that is not immediately visible but may contribute to the susceptibility to develop a particular behavior or syndrome. See the glossary, p. 84, for descriptions of other technical terms used in this article.) The goal of systems genetics is to provide an understanding of the complex relationship between the genome and disease by investigating intermediate biological processes. After investigating main effects, the first step in a systems genetics approach, as described here, is to search for gene-gene (i.e., epistatic) reactions.

Synergistic effect of the genetic polymorphisms of the renin-angiotensin-aldosterone system on high-altitude pulmonary edema: a study from Qinghai-Tibet altitude

The pathogenesis of high-altitude pulmonary edema (HAPE) has been at least partially attributed to the local dysregulation of the renin-angiotensin-aldosterone system (RAAS) cascade. To address this issue, we conducted the largest nested case-control study to-date to explore the association between variations in RAAS genes and HAPE in Chinese population. We recruited 140 HAPE patients and 144 controls during the construction of Qinghai-Tibet railway and genotyped 10 gene polymorphisms evenly interspersed in 5 RAAS candidate genes. The data were analyzed by haplotype and multifactor dimensionality reduction (MDR). The single-locus analysis showed that CYP11B2 C-344T and K173R and ACE A-240T polymorphisms were significantly associated with HAPE after Bonferroni correction (P<0.005). The linkage analysis constructed a linkage block including C-344T and K173R polymorphisms in complete linkage disequilibrium with each other, while occurred with significantly different frequencies between HAPE and control groups. The gene-gene interaction analysis found the overall best model including ACE A-240T and A2350G and CYP11B2 C-344T polymorphisms with strong synergistic effect. This model had a maximum testing accuracy of 68.61% and a maximum cross validation consistency of 9 out of 10 (P=0.004). The homozygous genotype combination of -240AA, 2350GG and -344TT conferred high genetic susceptibility to HAPE, which was further strengthened by haplotype analysis. Our results add evidence for synergistic effect of RAAS gene polymorphisms on HAPE susceptibility. Moreover, we proposed a promising data-mining analytical approach (MDR) for detecting and characterizing gene-gene interactions.

Particulate air pollution, oxidative stress genes, and heart rate variability in an elderly cohort

BACKGROUND AND OBJECTIVES

We have previously shown that reduced defenses against oxidative stress due to glutathione S-transferase M1 (GSTM1) deletion modify the effects of PM(2.5) (fine-particulate air pollution of < 2.5 microm in aerodynamic diameter) on heart rate variability (HRV) in a cross-sectional analysis of the Normative Aging Study, an elderly cohort. We have extended this to include a longitudinal analysis with more subjects and examination of the GT short tandem repeat polymorphism in the heme oxygenase-1 (HMOX-1) promoter.

METHODS

HRV measurements were taken on 539 subjects. Linear mixed effects models were fit for the logarithm of HRV metrics-including standard deviation of normal-to-normal intervals (SDNN), high frequency (HF), and low frequency (LF)-and PM(2.5) concentrations in the 48 hr preceding HRV measurement, controlling for confounders and a random subject effect.

RESULTS

PM(2.5) was significantly associated with SDNN (p = 0.04) and HF (p = 0.03) in all subjects. There was no association in subjects with GSTM1, whereas there was a significant association with SDNN, HF, and LF in subjects with the deletion. Similarly, there was no association with any HRV measure in subjects with the short repeat variant of HMOX-1, and significant associations in subjects with any long repeat. We found a significant three-way interaction of PM(2.5) with GSTM1 and HMOX-1 determining SDNN (p = 0.008), HF (p = 0.01) and LF (p = 0.04). In subjects with the GSTM1 deletion and the HMOX-1 long repeat, SDNN decreased by 13% [95% confidence interval (CI), -21% to -4%], HF decreased by 28% (95% CI, -43% to -9%), and LF decreased by 20% (95% CI, -35% to -3%) per 10 microg/m(3) increase in PM.

CONCLUSIONS

Oxidative stress is an important pathway for the autonomic effects of particles.

A multi-marker test based on family data in genome-wide association study

BACKGROUND

Complex diseases are believed to be the results of many genes and environmental factors. Hence, multi-marker methods that can use the information of markers from different genes are appropriate for mapping complex disease genes. There already have been several multi-marker methods proposed for case-control studies. In this article, we propose a multi-marker test called a Multi-marker Pedigree Disequilibrium Test (MPDT) to analyze family data from genome-wide association studies. If the parental phenotypes are available, we also propose a two-stage test in which a genomic screening test is used to select SNPs, and then the MPDT is used to test the association of the selected SNPs.

RESULTS

We use simulation studies to evaluate the performance of the MPDT and the two-stage approach. The results show that the MPDT constantly outperforms the single marker transmission/disequilibrium test (TDT) 1. Comparing the power of the two-stage approach with that of the one-stage approach, which approach is more powerful depends on the value of the prevalence; when the prevalence is no less than 10%, the two-stage approach may be more powerful than the one-stage approach. Otherwise, the one-stage approach is more powerful.

CONCLUSION

The proposed MPDT, is more powerful than the single marker TDT. When the parental phenotypes are available and the prevalence is no less than 10%, the proposed two-stage approach is more powerful than the one-stage approach.

No evidence for association between 19 cholinergic genes and bipolar disorder

Cholinergic dysfunction has been proposed for the pathogenesis of bipolar disorder (BD), and we have therefore performed a systematic association study of cholinergic system genes in BD (including schizoaffective disorder bipolar type). We genotyped 93 single nucleotide polymorphisms (SNPs) in 19 genes (CHAT, CHRM1-5, CHRNA1-7, CHRNA9, CHRNA10, and CHRNB1-4) in two series of samples: the National Institute of Mental Health (NIMH) Genetics Initiative pedigrees with 474 samples from 152 families, and the Clinical Neurogenetics (CNG) pedigrees with 83 samples from 22 multiplex families. Sib-transmission/disequilibrium test (sib_TDT) analysis showed nominally significant transmission bias for four SNPs (CHRNA2: rs7017417, P = 0.024; CHRNA5: rs514743, P = 0.031; CHRNB1: rs2302762, P = 0.049; CHRNB4: rs1948, P = 0.031). Haploview analyses showed nominally significant transmission bias of several haplotypes in CHRNA2, CHRNA7, CHRNB1, and CHRNB4, respectively. However, none of these associations reached gene-wide significance after correction by permutation. Alcohol dependence (including alcohol abuse) was not a significant covariate in the present genetic association analysis. Thus, it is unlikely that these 19 cholinergic genes play a major role in the pre-disposition to BD in these pedigrees.

Novel methods for detecting epistasis in pharmacogenomics studies

The importance of gene–gene and gene–environment interactions in the underlying genetic architecture of common, complex phenotypes is gaining wide recognition in the field of pharmacogenomics. In epidemiological approaches to mapping genetic variants that predict drug response, it is important that researchers investigate potential epistatic interactions. In the current review, we discuss data-mining tools available in genetic epidemiology to detect such interactions and appropriate applications. We survey several classes of novel methods available and present an organized collection of successful applications in the literature. Finally, we provide guidance as to how to incorporate these novel methods into a genetic analysis. The overall goal of this paper is to aid researchers in developing an analysis plan that accounts for gene–gene and gene–environment in their own work.

A balanced accuracy function for epistasis modeling in imbalanced datasets using multifactor dimensionality reduction

Multifactor dimensionality reduction (MDR) was developed as a method for detecting statistical patterns of epistasis. The overall goal of MDR is to change the representation space of the data to make interactions easier to detect. It is well known that machine learning methods may not provide robust models when the class variable (e.g. case-control status) is imbalanced and accuracy is used as the fitness measure. This is because most methods learn patterns that are relevant for the larger of the two classes. The goal of this study was to evaluate three different strategies for improving the power of MDR to detect epistasis in imbalanced datasets. The methods evaluated were: (1) over-sampling that resamples with replacement the smaller class until the data are balanced, (2) under-sampling that randomly removes subjects from the larger class until the data are balanced, and (3) balanced accuracy [(sensitivity+specificity)/2] as the fitness function with and without an adjusted threshold. These three methods were compared using simulated data with two-locus epistatic interactions of varying heritability (0.01, 0.025, 0.05, 0.1, 0.2, 0.3, 0.4) and minor allele frequency (0.2, 0.4) that were embedded in 100 replicate datasets of varying sample sizes (400, 800, 1600). Each dataset was generated with different ratios of cases to controls (1 : 1, 1 : 2, 1 : 4). We found that the balanced accuracy function with an adjusted threshold significantly outperformed both over-sampling and under-sampling and fully recovered the power. These results suggest that balanced accuracy should be used instead of accuracy for the MDR analysis of epistasis in imbalanced datasets.

Interaction between variants in the interleukin-4 receptor alpha and interleukin-9 receptor genes in childhood wheezing: evidence from a birth cohort study

BACKGROUND

Several polymorphisms in the IL-4 receptor alpha (IL4RA) gene have been associated with asthma and atopy, but with variable success in different populations. Immunologic studies suggest that IL4RA may interact with other cytokines and receptors, and gene-gene interactions have also been observed with respect to asthma. Such interactions have been proposed to explain partly the difficulties in replicating association studies.

METHODS

Using the prospective birth cohort BAMSE, we examined eight single nucleotide polymorphisms (SNPs) and corresponding haplotypes in the IL4RA gene in relation to wheezing and sensitization up to age 4. We also evaluated potential interaction effects (departure from a multiplicative interaction model) between the IL4RA SNPs and four SNPs in the IL-9 receptor (IL9R) gene previously associated with childhood wheezing.

RESULTS

We found no main effect of the IL4RA SNPs alone and only weak associations to wheezing and sensitization when haplotypes were considered. Gene-gene interactions between several IL4RA and IL9R SNPs with regard to wheezing were observed (P=0.009), especially between IL4RA Q576R (rs1801275) and IL9R rs731476 (P=0.005). An interaction was also seen between IL4RA and IL9R haplotypes.

CONCLUSION

Variants in the IL4RA gene alone may not exert any major influence on susceptibility to asthma-related diseases in childhood, but in combination with other genes, such as IL9R, IL4RA may be an important gene for disease susceptibility.

A partially linear tree-based regression model for assessing complex joint gene-gene and gene-environment effects

The success of genetic dissection of complex diseases may greatly benefit from judicious exploration of joint gene effects, which, in turn, critically depends on the power of statistical tools. Standard regression models are convenient for assessing main effects and low-order gene-gene interactions but not for exploring complex higher-order interactions. Tree-based methodology is an attractive alternative for disentangling possible interactions, but it has difficulty in modeling additive main effects. This work proposes a new class of semiparametric regression models, termed partially linear tree-based regression (PLTR) models, which exhibit the advantages of both generalized linear regression and tree models. A PLTR model quantifies joint effects of genes and other risk factors by a combination of linear main effects and a non-parametric tree -structure. We propose an iterative algorithm to fit the PLTR model, and a unified resampling approach for identifying and testing the significance of the optimal "pruned" tree nested within the tree resultant from the fitting algorithm. Simulation studies showed that the resampling procedure maintained the correct type I error rate. We applied the PLTR model to assess the association between biliary stone risk and 53 single nucleotide polymorphisms (SNPs) in the inflammation pathway in a population-based case-control study. The analysis yielded an interesting parsimonious summary of the joint effect of all SNPs. The proposed model is also useful for exploring gene-environment interactions and has broad implications for applying the tree methodology to genetic epidemiology research.

Symbolic modeling of epistasis

The workhorse of modern genetic analysis is the parametric linear model. The advantages of the linear modeling framework are many and include a mathematical understanding of the model fitting process and ease of interpretation. However, an important limitation is that linear models make assumptions about the nature of the data being modeled. This assumption may not be realistic for complex biological systems such as disease susceptibility where nonlinearities in the genotype to phenotype mapping relationship that result from epistasis, plastic reaction norms, locus heterogeneity, and phenocopy, for example, are the norm rather than the exception. We have previously developed a flexible modeling approach called symbolic discriminant analysis (SDA) that makes no assumptions about the patterns in the data. Rather, SDA lets the data dictate the size, shape, and complexity of a symbolic discriminant function that could include any set of mathematical functions from a list of candidates supplied by the user. Here, we outline a new five step process for symbolic model discovery that uses genetic programming (GP) for coarse-grained stochastic searching, experimental design for parameter optimization, graphical modeling for generating expert knowledge, and estimation of distribution algorithms for fine-grained stochastic searching. Finally, we introduce function mapping as a new method for interpreting symbolic discriminant functions. We show that function mapping when combined with measures of interaction information facilitates statistical interpretation by providing a graphical approach to decomposing complex models to highlight synergistic, redundant, and independent effects of polymorphisms and their composite functions. We illustrate this five step SDA modeling process with a real case-control dataset.

Exploiting gene-environment interaction to detect genetic associations

Complex disease by definition results from the interplay of genetic and environmental factors. However, it is currently unclear how gene-environment interaction can best be used to locate complex disease susceptibility loci, particularly in the context of studies where between 1,000 and 1,000,000 markers are scanned for association with disease. We present a joint test of marginal association and gene-environment interaction for case-control data. We compare the power and sample size requirements of this joint test to other analyses: the marginal test of genetic association, the standard test for gene-environment interaction based on logistic regression, and the case-only test for interaction that exploits gene-environment independence. Although for many penetrance models the joint test of genetic marginal effect and interaction is not the most powerful, it is nearly optimal across all penetrance models we considered. In particular, it generally has better power than the marginal test when the genetic effect is restricted to exposed subjects and much better power than the tests of gene-environment interaction when the genetic effect is not restricted to a particular exposure level. This makes the joint test an attractive tool for large-scale association scans where the true gene-environment interaction model is unknown.

Detection of gene x gene interactions in genome-wide association studies of human population data

Empirical evidence supporting the commonality of gene x gene interactions, coupled with frequent failure to replicate results from previous association studies, has prompted statisticians to develop methods to handle this important subject. Nonparametric methods have generated intense interest because of their capacity to handle high-dimensional data. Genome-wide association analysis of large-scale SNP data is challenging mathematically and computationally. In this paper, we describe major issues and questions arising from this challenge, along with methodological implications. Data reduction and pattern recognition methods seem to be the new frontiers in efforts to detect gene x gene interactions comprehensively. Currently, there is no single method that is recognized as the 'best' for detecting, characterizing, and interpreting gene x gene interactions. Instead, a combination of approaches with the aim of balancing their specific strengths may be the optimal approach to investigate gene x gene interactions in human data.

Genetic epidemiologic studies of preterm birth: guidelines for research

Over the last decade, it has become increasingly apparent that the cause of preterm birth is multifactorial, involving both genetic and environmental factors. With the development of new technologies capable of probing the genome, exciting possibilities now present themselves to gain new insight into the mechanisms leading to preterm birth. This review aims to develop research guidelines for the conduct of genetic epidemiology studies of preterm birth with the expectation that this will ultimately facilitate the comparison of data sets between study cohorts, both nationally and internationally. Specifically, the 4 areas addressed in this review includes: (1) phenotypic criteria, (2) study design, (3) considerations in the selection of control populations, and (4) candidate gene selection. This article is the product of discussions initiated by the authors at the 3rd International Workshop on Biomarkers and Preterm Birth held at the University of California, Los Angeles, Los Angeles, CA, in March 2005.

Two-stage two-locus models in genome-wide association

Studies in model organisms suggest that epistasis may play an important role in the etiology of complex diseases and traits in humans. With the era of large-scale genome-wide association studies fast approaching, it is important to quantify whether it will be possible to detect interacting loci using realistic sample sizes in humans and to what extent undetected epistasis will adversely affect power to detect association when single-locus approaches are employed. We therefore investigated the power to detect association for an extensive range of two-locus quantitative trait models that incorporated varying degrees of epistasis. We compared the power to detect association using a single-locus model that ignored interaction effects, a full two-locus model that allowed for interactions, and, most important, two two-stage strategies whereby a subset of loci initially identified using single-locus tests were analyzed using the full two-locus model. Despite the penalty introduced by multiple testing, fitting the full two-locus model performed better than single-locus tests for many of the situations considered, particularly when compared with attempts to detect both individual loci. Using a two-stage strategy reduced the computational burden associated with performing an exhaustive two-locus search across the genome but was not as powerful as the exhaustive search when loci interacted. Two-stage approaches also increased the risk of missing interacting loci that contributed little effect at the margins. Based on our extensive simulations, our results suggest that an exhaustive search involving all pairwise combinations of markers across the genome might provide a useful complement to single-locus scans in identifying interacting loci that contribute to moderate proportions of the phenotypic variance.

Although there is growing appreciation that attempting to map genetic interactions in humans may be a fruitful endeavor, there is no consensus as to the best strategy for their detection, particularly in the case of genome-wide association where the number of potential comparisons is enormous. In this article, the authors compare the performance of four different search strategies to detect loci which interact in genome-wide association—a single-locus search, an exhaustive two-locus search, and two, two-stage procedures in which a subset of loci initially identified with single-locus tests are analyzed using a full two-locus model. Their results show that when loci interact, an exhaustive two-locus search across the genome is superior to a two-stage strategy, and in many situations can identify loci which would not have been identified solely using a single-locus search. Their findings suggest that an exhaustive search involving all pairwise combinations of markers across the genome may provide a useful complement to single-locus scans in identifying interacting loci that contribute to moderate proportions of the phenotypic variance.

Powerful multilocus tests of genetic association in the presence of gene-gene and gene-environment interactions

In modern genetic epidemiology studies, the association between the disease and a genomic region, such as a candidate gene, is often investigated using multiple SNPs. We propose a multilocus test of genetic association that can account for genetic effects that might be modified by variants in other genes or by environmental factors. We consider use of the venerable and parsimonious Tukey's 1-degree-of-freedom model of interaction, which is natural when individual SNPs within a gene are associated with disease through a common biological mechanism; in contrast, many standard regression models are designed as if each SNP has unique functional significance. On the basis of Tukey's model, we propose a novel but computationally simple generalized test of association that can simultaneously capture both the main effects of the variants within a genomic region and their interactions with the variants in another region or with an environmental exposure. We compared performance of our method with that of two standard tests of association, one ignoring gene-gene/gene-environment interactions and the other based on a saturated model of interactions. We demonstrate major power advantages of our method both in analysis of data from a case-control study of the association between colorectal adenoma and DNA variants in the NAT2 genomic region, which are well known to be related to a common biological phenotype, and under different models of gene-gene interactions with use of simulated data.

Genetic associations in preterm birth: a primer of marker selection, study design, and data analysis

Spontaneous preterm birth (PTB; delivery before 37 weeks gestation) is a primary risk factor for infant morbidity and mortality. The etiology is unclear, but there is evidence that there is a genetic predisposition to PTB. Armed with the suggestion of genetic risk factors and the failure to identify useful biomarkers, investigators are starting to actively pursue the role of genetic predisposition in PTB. Several studies have been done to date assessing the role of single gene variants. However, positive findings have failed to replicate. We argue that heterogeneity in study designs, definition of phenotype, single-nucleotide polymorphism (SNP) selection, population selection, and sample size makes data interpretation difficult in complex phenotypes such as PTB. In this review, we introduce general concepts of study designs in genetic epidemiology, selection of candidate genes and markers for analysis, and analytical methodologies. We also introduce how the concept of gene-gene interactions (biologic epistasis) and gene-environment interactions may affect the predisposition to PTB.

A tutorial on statistical methods for population association studies

Although genetic association studies have been with us for many years, even for the simplest analyses there is little consensus on the most appropriate statistical procedures. Here I give an overview of statistical approaches to population association studies, including preliminary analyses (Hardy-Weinberg equilibrium testing, inference of phase and missing data, and SNP tagging), and single-SNP and multipoint tests for association. My goal is to outline the key methods with a brief discussion of problems (population structure and multiple testing), avenues for solutions and some ongoing developments.

GAIA: an easy-to-use web-based application for interaction analysis of case-control data

Background

The advent of cheap, large scale genotyping has led to widespread adoption of genetic association mapping as the tool of choice in the search for loci underlying susceptibility to common complex disease. Whilst simple single locus analysis is relatively trivial to conduct, this is not true of more complex analysis such as those involving interactions between loci. The importance of testing for interactions between loci in association analysis has been highlighted in a number of recent high profile publications.

Results

Genetic Association Interaction Analysis (GAIA) is a web-based application for testing for statistical interactions between loci. It is based upon the widely used case-control study design for genetic association analysis and is designed so that non-specialists may routinely apply tests for interaction. GAIA allows simple testing of both additive and additive plus dominance interaction models and includes permutation testing to appropriately correct for multiple testing. The application will find use both in candidate gene based studies and in genome-wide association studies. For large scale studies GAIA includes a screening approach which prioritizes loci (based on the significance of main effects at one or both loci) for further interaction analysis.

Conclusion

GAIA is available at

Associations of tumor necrosis factor G-308A with childhood asthma and wheezing

RATIONALE

Tumor necrosis factor (TNF) mediates a spectrum of airway inflammatory responses, including those to air pollutants, and is an asthma candidate gene. One TNF promoter variant (G-308A) affects expression of TNF and has been associated with inflammatory diseases; however, studies of asthma have been inconsistent. Because ozone produces oxidative stress, increased airway TNF, and inflammation, the associations of the -308 TNF polymorphism with asthma may vary by ozone exposure and variants of oxidant defense genes glutathione-S-transferase (GST) M1 and GSTP1.

OBJECTIVES

To investigate the association of TNF G-308A with asthma and wheezing and to determine whether these associations vary with ozone exposure and GSTM1 and GSTP1 genotype.

METHODS

We studied associations of TNF-308 genotype with lifetime and current wheezing and asthma among 3,699 children in the Children's Health Study. We examined differences in associations with community ozone and by GSTM1 null and GSTP1 105 Ile/Val (A105G) genotype.

RESULTS

Children with TNF-308 GG had decreased risk of asthma (odds ratio, 0.8; 95% confidence interval, 0.7-0.9) and lifetime wheezing (odds ratio, 0.8; 95% confidence interval, 0.7-0.9). The protective effects of GG genotype on wheezing outcomes were of greater magnitude in lower compared with higher ozone communities. These findings were replicated in the two cohorts of fourth-grade children recruited in 1993 and 1996. The reduction of the protective effect from the -308 GG genotype with higher ozone exposure was most marked in the GSTM1 null and GSTP1 Ile/Ile groups.

CONCLUSIONS

The TNF-308 GG genotype may have a protective role in asthma pathogenesis, depending on airway oxidative stress levels.