Statistical Methods in Genetic Studies

申请者的主要研究兴趣是开发有效的统计和计算方法来分析不同类型的数据和解决遗传研究中出现的问题。申请人还对传染病的传播建模感兴趣，重点是制定方法学，以改进个人一级模型的适合性，并解决数据中缺失信息的问题。因此，拟议的研究分为两个主题：遗传研究的统计方法和传染病模型。**在遗传学方面，高通量遗传数据为确定基因变异和感兴趣的特征之间的关系提供了机会。例如，在人类基因组中，数以百万计的单核苷酸多态(SNPs)可用于遗传学研究中的分析。然而，需要强大而稳健的统计方法和计算工具来分析如此海量的数据。因此，建议的研究将侧重于使用SNP数据进行遗传分析。**对多个性状的遗传关联进行联合分析，能够识别影响多个性状的常见遗传变异；在遗传学文献中，这被称为多效性效应。为了有效地调查疾病的发展，纵向队列研究旨在获得随着时间的推移个体内各种与疾病相关的特征的重复测量。与多个纵向性状的遗传关联的联合分析甚至更具挑战性，而且还没有分析相关个人数据的方法，例如家庭数据。拟议的研究旨在解决这些非常重要的问题。人们还知道，复杂的性状由多个(2个或更多)基因决定。一些基因可能与其他基因相互作用，而其他基因可能与环境因素和时间(如年龄)相互作用。在这个拟议的研究计划中，将开发用于识别基因-环境交互作用、时变基因和基因-基因交互作用的方法，这对于阐明复杂性状的潜在机制至关重要。**最近，在许多实际情况下，将SNP基因型从小小组(低密度)输入到大小组(高密度)已被考虑。使用较大的小组进行统计分析可以显著提高研究的能力，但使用较小的小组可以显著降低基因分型成本。因此，研究人员感兴趣的是对较低密度的面板进行基因分型，并结合对未分型的SNP进行准确的归因法。从大面板中选择信息丰富的SNPs子集来设计低密度面板可以显著提高计算精度。待开发的推算方法考虑了表型信息，并可以极大地提高重要SNP的准确性，例如与感兴趣的性状相关的SNP。**申请人还对传染病传播的模型感兴趣。在传染病数据中，遗漏信息很常见，例如未观察到或部分观察到的接触网络和未观察到的感染期。在传染病建模领域拟议的研究将侧重于开发方法，以改进个人一级模型的适合性，并解决数据中缺失信息的问题。**拟议工作的影响将在涉及人类遗传学、统计遗传学、动物育种的遗传改进以及人和动物的传染病的研究社区感受到。

The primary research interest of the applicant is on developing efficient statistical and computational methods for analyzing diverse types of data and addressing issues arising from genetic studies. The applicant is also interested in modeling the spread of infectious disease, with emphases on methodology development for improving the fit of individual-level models and addressing issues of missing information in data. The proposed research is therefore presented as two themes: Statistical methods for genetic studies and Infectious disease modelling.**In genetics, high-throughput genetic data provide opportunities for identifying relationships between genetic variants and traits of interest. In the human genome, for example, millions of single nucleotide polymorphisms (SNPs) are available for analysis in genetic studies. However, powerful and robust statistical methods and computational tools are needed to analyze such massive amounts of data. So, the proposed research will be focusing on genetic analysis using SNP data.**Joint analysis of genetic association with multiple traits enables the identification of common genetic variants that influence more than one trait; in genetics literature, this is known as the pleiotropic effect. To effectively investigate the development of a disease, longitudinal cohort studies are designed to obtain repeated measures of a variety of disease-related traits within an individual over time. Joint analysis of genetic association with multiple longitudinal traits is even more challenging, and methods for analyzing data from related individuals, such as family data, are not yet available. The proposed research aims to tackle these very important problems. It is also known that multiple (2 or more) genes are responsible for complex traits. Some genes might interact with other genes while others might interact with environmental factors and time (e.g., age). The methods to be developed in this proposed research program for identifying gene-environmental interactions, time-varying genes, and gene-gene interactions are critical for elucidating the underlying mechanism of a complex trait. **Recently, imputing SNP genotypes from a small panel (lower density) to a large panel (higher density) has been considered in many practical situations. Statistical analysis using a larger panel can significantly improve the power of the study but a small panel can substantially lower the genotyping cost. As such, researchers are interested in genotyping a lower density panel in conjunction with an accurate imputation method for the untyped SNPs. Selecting a subset of informative SNPs from a large panel to design low-density panels can substantially improve the imputation accuracy. The imputation method to be developed takes phenotype information into account and can greatly improve the accuracy for important SNPs, such as those associated with the trait of interest. **The applicant is also interested in the modeling of the spread of infectious disease. In infectious disease data, missing information, such as the unobserved or partially observed contact network and the unobserved infectious period, is common. The proposed research in the area of infectious disease modeling will focus on developing methodologies for improving the fit of individual-level models and addressing issues of missing information in data. **The impact of the proposed work will be felt in research communities concerned with human genetics, statistical genetics, genetic improvement of animal breeding, and infectious diseases in humans and animals.