New Statistical Methods for High-Dimensional Association Tests with Applications to Large-Scale Genetic Data

该研究通过开发和实施强大且理论上合理的统计方法来检测强大的生物标志物关联，解决了大规模生物医学研究中的几个挑战。所开发的方法受到大规模遗传关联研究的推动并将对其产生重大影响，并且也广泛适用于其他学科（例如调查抽样和心理健康成像研究）。所开发的方法将用于检测与多种心脏代谢特征相关的新型遗传生物标志物，并为生物学和临床研究产生新的假设。该项目还将解决一些长期存在的统计问题，并将提供理论上健全和比常用的更强大的方法。项目团队将整合研究成果，培养快速发展的生物医学数据科学领域的下一代本科生和研究生。该项目还将促进教学、培训和学习，并扩大代表性不足群体学生的参与。最近的方法学和计算方面的进步促进了统计方法在该领域大规模全基因组关联研究中分析简单（主要是单一）疾病结果的应用。然而，这些研究只确定了一小部分风险变异，可能还有更多的普通变异和/或罕见变异尚未被发现。现有的方法和统计理论不足以分析与聚类结果（如纵向结果）和多个相关和/或次要结果的高维关联。本项目旨在解决这一迫切需求，开发具有坚实理论基础的新的统计方法，整合多种相关表型，以识别复杂性状的新遗传变异。特别是，该项目将(1)开发强大的统计方法，用于测试与聚类结果的高维关联；(2)建立理论健全、功能强大的多次要性状关联检验统计方法；(3)开发并应用统一的建模框架，该框架应用我们开发的统计方法，利用全基因组测序数据，并集成功能注释数据，以帮助识别和剖析罕见变异在心脏代谢性状中的作用。该教育计划的目标是将统计遗传学的最新研究发展纳入现有/新课程，为学生将来从事生物医学/健康信息学的职业做好准备。研究还将包括实现这些方法的软件开发。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The research tackles several challenges in large-scale biomedical studies through the development and implementation of powerful and theoretically sound statistical methods to detect robust biomarker associations. The developed methods are motivated by and will have significant impact on large-scale genetic association studies, and are also broadly applicable to other disciplines (e.g. survey sampling and mental health imaging studies). The developed methods will be applied to detect novel genetic biomarkers that are associated with multiple cardiometabolic traits, and generate novel hypotheses for biological and clinical investigation. The project will also solve some long-standing problems in statistics, and will provide theoretically sound and much more powerful methods than the commonly used ones. The project team will integrate the research results into training the next generation undergraduate and graduate students in the fast growing field of biomedical data science. This project will also promote teaching, training and learning, and broaden the participation of students from under-represented groups.Recent methodological and computational advances have facilitated the applications of statistical methods to analyze simple (primarily single) disease outcomes in large-scale genome-wide association studies in the field. However, these studies have identified only a small proportion of the risk variants and there likely remain many more common variants with modest effect sizes and/or rare variants yet to be discovered. Existing methods and statistical theories are not adequate for analyzing high-dimensional association with clustered outcomes (e.g. longitudinal outcomes), and multiple correlated and/or secondary outcomes. This project aims to address this urgent need by developing new statistical methods with solid theoretical foundation to integrate multiple correlated phenotypes to identify novel genetic variants for complex traits. In particular, the project will (1) develop powerful statistical methods for testing high-dimensional association with clustered outcomes; (2) develop theoretically sound and powerful statistical methods for testing association with multiple secondary traits; and (3) develop and apply a unified modeling framework that applies our developed statistical methods, leverages the whole genome sequencing data, and integrates functional annotation data to help identify and dissect the role of rare variants on the cardiometabolic traits. The objective of the education plan is to integrate the latest research development in statistical genetics into existing/new courses to prepare students for their future professions in biomedical/health informatics. The research will also include software development to implement the methods.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.