Inferring Multiple-SNP Disease Association with DNA Resequence Data

利用 DNA 测序数据推断多 SNP 疾病关联

• 批准号: 7692992
• 负责人: Carlos Daniel Bustamante
• 金额: $ 38.03万
• 依托单位: CORNELL UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2008
• 资助国家: 美国
• 起止时间: 2008-09-28 至 2011-07-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/7692992
• 关键词: Accounting; Affect; Age; Alleles; Base Ratios; Bayesian Method; Blood Pressure; Body mass index; Candidate Disease Gene; Chronic; Cohort Studies; Collection; Complex; Computer Simulation; DNA Resequencing; DNA Sequence; Data; Data Set; Data Sources; Development; Disease; Disease Association; Disease Outcome; Disease susceptibility; Economic Inflation; Environment; Environmental Risk Factor; Equilibrium; Framingham Heart Study; Frequencies; Genes; Genetic Databases; Genetic Models; Genetic Population Study; Genetic Recombination; Genome; Genotype; Goals; Haplotypes; Human; Individual; Institutes; Maps; Markov Chains; Measures; Medicine; Methods; Metric; Modeling; Molecular; Movement; Mutation; Phase; Phenotype; Play; Population; Population Genetics; Probability; Property; Reading; Risk; Risk Factors; Role; Sampling; Simulate; Site; Specific qualifier value; Statistical Methods; Statistical Models; Stratification; Structure; Sum; Tail; Techniques; Testing; Transcript; Variant; base; case control; design; disorder risk; factor A; flexibility; gene environment interaction; gene function; genetic variant; genome wide association study; improved; interest; simulation; statistics; theories; trait

DESCRIPTION (provided by applicant): Whole-genome association testing by genotyping common SNPs has promise for identification of genetic variants that are causal to elevated risk of complex disorders, but there is pressing need to apply deep resequencing to the regions found by these studies to further understand the disease association. Full sequence data in a population sample will no longer suffer from problems of ascertainment bias, and the full spectrum of population genetic models may be fitted to the data. But there remains a serious challenge to identify optimal means of establishing association between sequence variants and disease risk, and we will pursue four aims toward this goal. Specific Aim 1 will consider the role of rare variants, starting with the challenge of calling singleton heterozygous sites in large samples and dealing with errors in these calls. Rare variants do not provide statistical power to be tested for phenotype associations individually, but a variety of tests of collective effects of rare variants are proposed. The basic idea of considering the collections of genotypic attributes that distinguish the phenotypic tails of the distribution of measured genotypes will be explored extensively. Specific Aim 2 will consider the case of dense resequencing of specific candidate genome regions in case-control and cohort studies. Both unphased multi-site genotype data and phased haplotypes inferred from the genotype data represent samples from a population that may or may not fit well-studied population genetic and demographic models. The causal model connecting phenotype to genotype likely includes an integration of effects of multiple SNPs, including possibly highly non-additive effects. We propose a likelihood ratio based method that calculates the probability of observing the phenotypic data under a specific genetic model that can account for the combined effect of multiple SNPs. Specific Aim 3 will extend the model of Aim 2 to a Bayesian setting, applying Markov Chain Monte Carlo techniques for sampling from the posterior distribution of effects. This will allow the test to be applied to much larger data sets, including resequencing of regions after genome-wide association studies. Specific Aim 4 develops an improved and flexible Bayesian association mapping approach that can integrate disparate sources of data (such as age, intermediate phenotype, environment, etc.) to estimate the inflation of risk of disease for single or combinations of genetic variants, environmental conditional, age or combinations of factors. For all four aims, approaches will be tested with sample resequencing data as well as genotypic data generated by simulation of the coalescent with recombination under realistic human demographic models. In both cases phenotypes will be specified by a variety of genetic models. Data from the Framingham Heart Study, from GAIN genome-wide studies, and from the Sanger Institute data on transcript abundance of HapMap samples will be used to test the methods. DNA resequencing of large samples of individuals from case/control and cohort studies will yield information about associations with disease risk only if the properties of statistical models that describe the causal associations between genotype and phenotype are fully explored. This project is designed to develop an analytical framework that uses the underlying structure of the genetic data (based on population genetic principles) to provide the maximum statistical power for inference of association when many SNPs within the gene may contribute a small, non-additive portion of the increased risk. These methods will also be extended to include prior information about molecular mechanisms of gene function, where available, as well as environmental contributions to disease risk.

描述（由申请人提供）：通过对常见SNP进行基因分型的全基因组关联检测有望识别导致复杂疾病风险升高的遗传变异，但迫切需要对这些研究发现的区域进行深度重测序，以进一步了解疾病关联。群体样本中的全序列数据将不再受到确定偏差的问题的困扰，并且群体遗传模型的全谱可以拟合到数据。但是，确定序列变异与疾病风险之间建立关联的最佳方法仍然是一个严峻的挑战，我们将朝着这个目标追求四个目标。具体目标1将考虑罕见变异的作用，从在大样本中识别单基因杂合位点的挑战开始，并处理这些识别中的错误。罕见变异不提供统计功效来单独检验表型关联，但提出了罕见变异的集体效应的各种检验。将广泛探讨考虑区分测量基因型分布的表型尾部的基因型属性集合的基本思想。具体目标2将考虑病例对照和队列研究中特定候选基因组区域的密集重测序情况。未定相的多位点基因型数据和从基因型数据推断的定相单倍型都代表来自可能适合或可能不适合充分研究的群体遗传和人口统计模型的群体的样品。连接表型与基因型的因果模型可能包括多个SNP的效应的整合，包括可能的高度非加性效应。我们提出了一种基于似然比的方法，该方法计算在特定遗传模型下观察表型数据的概率，该模型可以解释多个SNP的组合效应。具体目标3将目标2的模型扩展到贝叶斯设置，应用马尔可夫链蒙特卡罗技术从效应的后验分布中进行采样。这将使该测试能够应用于更大的数据集，包括在全基因组关联研究后对区域进行重新测序。具体目标4开发了一种改进的和灵活的贝叶斯关联映射方法，可以集成不同的数据源（如年龄，中间表型，环境等）。估计单一或遗传变异组合、环境条件、年龄或因素组合的疾病风险膨胀。对于所有四个目标，方法将进行测试，样本重测序数据以及基因型数据的模拟下的现实人类人口模型的重组合并。在这两种情况下，表型将由各种遗传模型指定。来自Frachial Heart Study的数据、来自GAIN全基因组研究的数据以及来自桑格研究所关于HapMap样本的转录本丰度的数据将用于测试该方法。只有充分探索了描述基因型和表型之间因果关系的统计模型的特性，才能对病例/对照和队列研究中的大样本个体进行DNA重测序，从而获得与疾病风险相关的信息。该项目旨在开发一个分析框架，该框架使用遗传数据的基本结构（基于群体遗传学原理），当基因内的许多SNP可能导致风险增加的一小部分时，为关联推断提供最大的统计功效。这些方法还将扩展到包括有关基因功能分子机制的先前信息，以及环境对疾病风险的贡献。