Statistical genetics of aging-related genomic and phenotypic change

为了帮助分析和理解受许多基因和环境因素影响的与衰老相关的“复杂”性状，我们建议开发三种统计算法用于全基因组基因分型和高通量测序研究的分析。我们提出的新统计方法提供了分析其他类型数据的方法，例如，来自测序的线粒体DNA（mtDNA）变体，或用于全基因组关联研究（GWAS）的X染色体上的变体和具有更复杂结构的数据（例如，多个相关性状）。为了测试这些算法，我们利用了SardiNIA项目的特殊功能（见年度报告AG 000675 -07），该项目收集了超过300个数量性状的纵向数据以及撒丁岛种群创始人的全基因组遗传数据。 为了分析线粒体DNA变异及其对衰老相关性状的可能影响，针对核DNA开发的基因型调用和分析程序是不够的，因为每个细胞有100- 10，000个线粒体DNA拷贝，这些拷贝可以在任何位点发生变化（异质性），因此可以在不同拷贝的任何位置上具有4个碱基中的每一个。我们已经开发出一种算法，该算法的目标是识别mtDNA中的变异;它结合了每个序列读取中每个碱基的测序错误率，并且可以灵活地允许所有个体中变异位点处的不同等位基因分数。我们的程序是进一步适应的圆形线粒体基因组，从线性染色体的主要区别，假设大多数映射算法。我们正在评估来自2,000名SardiNIA项目参与者的全基因组测序的淋巴细胞mtDNA序列中的同质性和异质性。迄今为止的结果提供有关mtDNA单倍型群和遗传的同源和异质性在撒丁岛的信息。正如预期的那样，母亲和她们的孩子基本上共享所有的同质性，但异质性的比例较小。总体异质性随着年龄的增长而增加，但到目前为止估计的斜率很小，在20岁至80岁之间平均增加1个异质性，次要等位基因分数阈值为4%。 为了利用相关性状之间的相关性，从而增加遗传研究的统计能力，我们正在开发一种方法来搜索对多个数量性状具有多效性效应的基因/变体。我们的方法将一组相关性状和一组SNP（例如，由基因边界定义）投影到各自的正交主成分中，以便我们能够使用新的汇总统计量联合测试性状和SNP之间的关联。由于正交性，关联的重要性可以使用零假设下的模拟来有效地评估，而不是更计算密集的排列。我们将我们的方法应用于SardiNIA项目数据，其中我们首先关注三种脂质性状HDL，LDL和甘油三酯，并使用RefSeq基因边界将SNP分组为基因单元。为了证明我们的方法能够识别与一种以上血脂性状相关的基因，我们使用了Teslovich等人发表的结果，Nature（2010），其中鉴定了95个血脂基因座，21个基因座与多种脂质性状相关。我们能够表明，我们的方法可以丰富那些多效性基因座：例如，当95个基因座通过我们的方法排名时，前20%的基因座包括原始研究中40%的多效性基因座。我们的方法能够检测多个性状和多个遗传变异之间的联合关联。当一个基因对多个性状有中度影响，而标准GWAS无法检测时，它将具有显着的优势。

To help to analyze and understand aging-related "complex" traits that are affected by many genes and environmental factors, we propose to develop three statistical algorithms for the analyses of genome-wide genotyping and high-throughput sequencing studies. Our proposed new statistical methods provide means to analyze additional types of data e.g., mitochondrial DNA (mtDNA) variants from sequencing, or variants on the X chromosome for genome-wide association studies (GWAS) and data with more complicated structures (e.g., multiple related traits). To test these algorithms, we take advantage of the special features of the SardiNIA project (see Annual Report AG000675-07), which has collected longitudinal data for >300 quantitative traits together with the whole-genome genetic data in the founder Sardinia population. To analyze mitochondrial DNA variation and its possible effects on aging-related traits, the genotype-calling and analytic programs developed for nuclear DNA are not adequate, because each cell has 100-10,000 mtDNA copies that can vary at any site (heteroplasmy), and can therefore have each of the 4 bases at any position in various copies. We have developed an algorithm that is targeted to identify variants in mtDNA; it incorporates the sequencing error rate of each base in each sequence read and is flexible to allow for different allele fractions at a variant site across all individuals. Our procedure is further adapted to the circular mitochondrial genome, a key difference from the linear chromosomes assumed by most mapping algorithms. We are assessing homoplasmies and heteroplasmies in mtDNA sequences of lymphocytes from whole-genome sequencing of 2,000 SardiNIA Project participants. The results to date provide information about mtDNA haplogroups and the inheritance of homo- and heteroplasmies in Sardinia. As expected, mothers and their children share essentially all homoplasmies but a lesser proportion of heteroplasmies. The overall heteroplasmy increases with age, but the slope is small in the estimates thus far, yielding an average increase of 1 heteroplasmy between ages 20 and 80 with the minor allele fraction threshold at 4%. To take advantage of correlations between related traits and hence to increase statistical power for genetic studies, we are developing a method to search for genes/variants that have pleiotropic effects on multiple quantitative traits. Our method projects a group of related traits and a set of SNPs (defined, for example, by gene boundaries) into their respective orthogonal principal components such that we are able to jointly test the association between traits and SNPs using a new summary statistic. Because of the orthogonality, the significance of association can be efficiently evaluated using simulations under the null hypothesis instead of more computationally intensive permutations. We apply our method to the SardiNIA project data, where we have first focused on three lipid traits HDL, LDL, and Triglycerides, and use RefSeq gene boundaries to group SNPs into gene units. To show that our method is able to identify genes that are associated with more than one blood lipid trait, we use results published in Teslovich et al., Nature (2010), in which 95 loci for blood lipids were identified and 21 loci were associated with multiple lipid traits. We are able to show that our method can enrich those pleiotropic loci: for example, when the 95 loci are ranked by our method, the top 20% loci include 40% of the pleiotropic loci in the original study. Our method is able to detect joint associations between multiple traits and multiple genetic variants. It will have significant advantages when a gene has moderate effects on multiple traits that a standard GWAS is unable to detect.