Regression Models for Data Integration in Genetics and Genomics

该研究计划的总体目标是开发统计工具，以解决人类生物学和复杂性状分析的基础和转化研究中的问题。我建议开发灵活的统计模型和方法的数据分析，整合多个数据源，考虑到复杂的依赖结构固有的分子生物学研究的遗传，分子和环境因素负责个体之间的差异。微阵列、全基因组关联平台和下一代测序等技术的出现允许对个体之间的全基因组分子和遗传变异进行询问，并有可能更好地描绘异质性的来源。大规模组学数据方法的开发通常是通过一次关注一个或两个数据维度来务实地进行的。占主导地位的范式通常进行单变量或单基因关联的测试，而不是专注于因素的组合。通过开发、评估和应用能够调查多个因素的回归方法，我们希望对潜在的生物学和人与人之间差异的来源有新的认识。 我建议解决三个具体的统计困难。这些障碍包括当一套完整的测量方法无法适用于每个人时，当感兴趣的特征在人群中的个体分布不均匀时，以及当潜在的解释因素数量非常大时。作为研究培训的一部分，生物统计学和统计学的研究生和博士后研究员将在开发我提出的问题的解决方案方面发挥不可或缺的作用。除了向其他统计研究人员传播研究结果外，我们还计划开发和提供对许多领域的调查人员有用的可访问软件。

The general objective of the research program is to develop statistical tools to address problems in basic and translational research in human biology and complex trait analysis. I propose to develop flexible statistical models and methods of data analysis that integrate multiple sources of data, taking into account the complex dependence structures inherent in molecular biologic studies of the genetic, molecular and environmental factors responsible for differences among individuals. The advent of technologies such as microarrays, genome-wide association platforms, and next generation sequencing allows interrogation of genome-wide molecular and genetic variation among individuals, and the potential to better delineate sources of heterogeneity. The development of methodology for large-scale ‘omics data has generally proceeded pragmatically by focusing on one or two data dimensions at a time. The dominant paradigms have typically performed tests of single-variant or single-gene associations rather than focusing on combinations of factors. By developing, evaluating and applying regression methods that have the capacity to investigate multiple factors, we expect to gain new insight into underlying biology and the sources of differences among people. There are three particular statistical difficulties that I propose to address. These include obstacles that arise when a complete set of measurements is not available on everyone, when the trait of interest is not evenly distributed across individuals in the population, and when the number of potential explanatory factors is very large. As part of their research training, graduate students and post-doctoral fellows in biostatistics and statistics will play an integral role in developing solutions to the problems I have posed. In addition to disseminating the results of the research to other statistical researchers, we plan to develop and provide accessible software useful to investigators in many fields.