Gene-environment interaction vs quantile-dependent penetrance of established SNPs

描述(申请人提供)：基因-环境相互作用可被定义为一种基因的表型表达被环境改变，例如，与不锻炼的人相比，锻炼者的FTO基因多态对体重指数(BMI)的影响较弱。然而，我们的初步分析表明，FTO多态的表型表达在瘦与超重的个体中减少。这意味着另一种解释(图1)。基于运动者比非运动者更瘦的事实，我们假设体育运动影响BMI，进而影响FTO基因表达，而不是运动直接影响FTO基因表达。我们对1800名受试者的脂蛋白和BMI的初步分析表明，一种基因型的表型表达通常随着表型的分位数而增加，即当表型的值相对于其在人群中的分布而言较高时。我们将这种依赖称为分位数相关的外显度。这种相关性与标准回归模型不同，标准回归模型假定因变量和自变量(例如，表型与基因型)之间的相同关系适用于因变量的所有分位数。我们建议将分位数回归应用于通过NHLBI候选基因关联资源(CARE)、DBGaP和其他研究获得的数据，以评估分位数依赖的外显性是否适用于大多数其他的基因-表型关系。尽管我们的初步分析缺乏评估单个SNPs这一现象的统计能力，但它在遗传风险评分中的证明表明，SNP的大多数效应也必须是分位数相关的。我们还将测试先前关于基因-环境交互作用的断言是否可归因于分位数依赖的外显性，是否允许基因表达随性状分布的百分位数变化显著增加所解释的表型差异，以及分位数依赖的外显性是否可以扩展到SNP-SNP交互作用。这一建议是假说驱动的；也就是说，我们假设大多数基因-表型关联随着表型的百分位数的增加而显著增加。这一假设是基于这样一个前提，即最重要的基因-环境相互作用涉及到个体自身的生理环境，在这个环境中基因得到了表达。性状分布的最低百分位数到最高百分位数代表了一系列生理参数、基因组成和基因-基因相互作用，这些参数的存在可能是遗传变异表达的关键。据我们所知，分位数依赖的外显性并没有被认为是基因-表型关系的主要基础，也没有被认为是基因-环境相互作用的替代。 公共卫生相关性：我们之前已经表明，在总胆固醇、甘油三酯、高密度脂蛋白胆固醇和体重指数的遗传风险分数中，随着性状分布的百分位数的增加，基因对表型的影响也会增加。这一现象在一项对1800名受试者的研究的基线数据中得到了证实。这项提议的目的是在更大、更多样化的数据集上证明这一现象，并将研究结果推广到其他变量。我们认为，这一现象可能是表型-基因型关系的一般原理。这些结果可以更准确地定义基因-环境互作：a)环境对基因的影响，影响其表型表达；b)环境对表型的影响，影响基因的外显。

DESCRIPTION (provided by applicant): Gene-environment interactions may be defined as a genotype's phenotypic expression being altered by the environment, e.g., the weaker effect of FTO polymorphisms on body mass index (BMI) in exercisers compared to non-exercisers. However, our preliminary analyses suggest that the phenotypic expressions of FTO polymorphisms are diminished in lean vis-a-vis overweight individuals. This suggests an alternative interpretation (Figure 1). Based on the fact that exercisers are leaner than non-exercisers, we hypothesize that physical activity affects BMI, which in turn affects FTO gene expression, rather than exercise affecting FTO gene expression directly. Our preliminary analyses of lipoproteins and BMI in 1800 subjects suggest that a genotype's phenotypic expression often increases with the quantile of the phenotype, i.e., when the value of the phenotype is high relative to its distribution in the population. We refer to this dependence as quantile- dependent penetrance. This dependence differs from the standard regression model, which assumes that the same relationship between the dependent and independent variables (e.g., phenotype vs. genotype) applies to all quantiles of the dependent variable. We propose to apply quantile regression to data available through the NHLBI Candidate-Gene Association Resource (CARe), DBGaP, and other studies to assess whether quantile-dependent penetrance applies to most other genotype-phenotype relationships. Although our preliminary analyses lacked the statistical power to assess this phenomenon for individual SNPs, its demonstration in genetic risk scores suggests that the majority of SNP effects must also be quantile dependent. We will also test whether prior assertions of gene- environment interactions are attributable to quantile-dependent penetrance, whether allowing the genotypic expression to vary with the percentile of the trait distribution significantly increases the phenotypic variances explained, and whether quantile-dependent penetrance can be extended to SNP-SNP interactions. This proposal is hypothesis driven; i.e., we hypothesize that most genotype-phenotype associations increase substantially with the percentile of the phenotype. This hypothesis is based upon the premise that the most important gene-environment interaction involves an individual's own physiological environment within which the genes are expressed. The lowest to highest percentiles of a trait's distribution represent a range of physiologic parameters, genetic make-ups, and gene-gene interactions whose presence may be essential for the genetic variant to be expressed. To our knowledge, quantile-dependent penetrance has not been proposed as a primary basis for genotype-phenotype relationships, or as an alternative to gene-environment interactions. PUBLIC HEALTH RELEVANCE: We have previously shown that the effect of the genotype on a phenotype increases with the percentile of the trait distribution in genetic risk scores for total cholesterol, triglycerides, high-density lipoprotein cholesterol, and body mass index. This phenomenon was demonstrated using baseline data for a study of 1800 subjects. The purpose of this proposal is to demonstrate this phenomenon in much larger, more diverse data sets and to extend the findings to other variables. We believe that this phenomenon could be a general principal of phenotype-genotype relationships. The results could more sharply define gene-environment interactions into: a) effects of the environment on the genotype, which affects its phenotype expression, and b) effects of the environment on the phenotype which affects the penetrance of the genotype.