Analysis of quantitative genetic traits in a huge data set

数量性状变异和协变的遗传基础是人类遗传学、进化生物学和动植物育种的核心。在医学遗传学中，许多疾病，包括精神分裂症、心脏病和癌症，都是具有连续表型和易感性的复杂特征，具有导致遗传差异的多个基因组变异。在进化生物学中，适合性在很大程度上取决于这样的数量性状(例如，繁殖力、寿命)。在动植物育种中，大多数重要的经济性状都是数量性状(如牛奶、肉类和谷物产量、环境足迹、繁殖力)。统计基因组学需要巨大的数据集，因为许多变量(可能是数千个)可以聚集在一起，对任何单个数量性状都有贡献，它们的影响可以以复杂的方式组合(加性、显性、上位性)。此外，数量性状的遗传方差的重要部分由罕见的、效应大小较小或与其他变量高度相关的变量控制。只有在非常庞大的数据集中使用非常强大的统计模型时，才能分离这些数量性状变异的影响。我们将通过创建和分析一个包含来自世界上最大的商业育种计划的325000头猪的基因组序列、系谱和特征记录的数据库，在分子水平上分析25个数量性状的遗传基础。数据集将在我们为支持育种计划而开发的那些算法的基础上使用补偿和分析算法来创建和分析。数据集的大小和数据的质量将使我们能够解决三大问题：-1。哪些基因组变异控制哪些数量性状，它们如何控制它们，以及控制单个性状的多个变异如何相互作用？2.什么类型的机制导致性状共变？多效性和连锁不平衡在多大程度上有贡献？基因组区域对性状的联合作用的大小和符号的分布是什么？3.庞大的数据集在多大程度上帮助我们解决这些问题？我们第一次拥有了以低成本为数十万个个体生成基因组序列数据的技术，以及存储和分析这些数据的计算机能力。该项目的目标是从一个耗资150亿美元的猪育种计划中获得科学利益。我们之前的项目是问统计基因组学如何帮助动物育种；这个项目问的是动物育种如何帮助统计基因组学。

The genetic basis of quantitative trait variation and covariation is central to human genetics, evolutionary biology, and plant and animal breeding. In medical genetics many diseases, including schizophrenia, heart disease and cancer, are complex traits with continuous phenotypes and liabilities, which have multiple genome variants contributing genetic variance. In evolutionary biology fitness is largely due to such quantitative traits (e.g., fecundity, longevity). In plant and animal breeding most of the economically important traits are quantitative traits (e.g., milk, meat, and grain yields, environmental footprint, fecundity). Huge datasets are needed for statistical genomics because many variants (probably thousands), which can be clustered together, contribute to any individual quantitative trait and their effects can combine in complex ways (additive, dominant, epistatic). Moreover, important portions of the genetic variance of quantitative traits are controlled by variants that are rare, have small effect sizes or are highly correlated with other variants. The effects of such quantitative trait variants can only be separated when very powerful statistical models are used in very large data sets.We will analyse the genetic basis of 25 quantitative traits at the molecular level by creating and analysing a dataset containing genome sequences, pedigrees and trait records of 325000 pigs from the world's biggest commercial breeding programme. The dataset will be created and analysed using imputation and analysis algorithms based on those that we developed to support the breeding programme.The size of the dataset and the quality of the data will allow us to address three big questions:-1. Which genome variants control which quantitative traits, how do they control them and how do the multiple variants that control a single trait interact?2. What kinds of mechanisms cause traits to co-vary? To what extent does pleiotropy and linkage disequilibrium contribute? What is the distribution of the magnitude and sign of joint effects of genomic regions on pairs of traits? 3. To what extent do huge data sets help us address these questions? For the first time we have the technology to generate genome sequence data for hundreds of thousands of individuals at low cost and the computer power to store and analyse such data.The aim of this project is to harvest scientific benefits from a 15 year billion dollar pig breeding program. Our previous projects asked how statistical genomics helps animal breeding; this project asks how animal breeding helps statistical genomics.