Statistical Models for Genetic Studies, Using Network and Integrative Analysis

全基因组关联研究（GWAS）已经确定了数万种与基因组相关的遗传变异。 有数百种表型和疾病，在某些情况下提供了临床和医疗益处， 新的生物标志物和治疗靶点。然而，对复杂性状的研究往往 由于多源性、高维性和中等样本量，统计功效有限。而 招募患者以获得足够的样本量来识别所有相关的 遗传变异，我们最近表明，统计能力，以确定风险相关的遗传变异，可以 1）考虑到多种表型之间共有的遗传基础，即多效性， 以及2）合并基因组和遗传注释数据。然而，这些数据集的有效整合 随着遗传研究和注释数据数量的增加，在统计上变得更具挑战性。 本提案的目的是开发统计方法和软件，以改进识别和 风险变异的解释和促进对表型之间遗传关系的理解。这 将通过实现四个具体目标来实现这一目标。在目标1中，我们将开发一个贝叶斯图模型 通过整合多个GWAS数据集， 注释数据。在目标2中，我们将开发一个贝叶斯图模型，从 生物医学文献在目标3中，我们将开发一种统计方法来构建元注释， 有效地概括高维注释数据而不损失可解释性。在目标4中，我们将应用 这些方法用于非裔美国人血管并发症和自身免疫性疾病的遗传学研究 人口，PubMed文献和各种注释数据集。该研究具有创新性 因为它提出了一个新的统计框架，整合了多个GWAS，生物医学文献， 注释数据集，以改善风险变体的识别和解释。拟议的研究是 重要的是，它有望帮助改善疾病的诊断和治疗， 确定风险变异，并加强对疾病共同病因的了解。

Genome-wide association studies (GWAS) have identified tens of thousands of genetic variants associated with hundreds of phenotypes and diseases, which in some cases have provided clinical and medical benefits to patients with novel biomarkers and therapeutic targets. However, investigation of complex traits often suffers from limited statistical power due to polygenicity, high dimensionality, and moderate sample size. While it is practically challenging and costly to recruit patients to attain sufficient sample size to identify all associated genetic variants, we recently showed that statistical power to identify risk associated genetic variants can be significantly increased by 1) considering genetic basis shared among multiple phenotypes, namely pleiotropy, and 2) incorporating genomic and genetic annotation data. However, effective integration of these datasets becomes statistically more challenging as the number of genetic studies and annotation data increases. The objective of this proposal is to develop statistical methods and software to improve identification and interpretation of risk variants and to promote understanding of genetic relationship among phenotypes. This objective will be attained by pursuing four specific aims. In Aim 1, we will develop a Bayesian graphical model to identify risk variants and construct a phenotype network, by integrating multiple GWAS datasets with various annotation data. In Aim 2, we will develop a Bayesian graphical model to build a phenotype network from biomedical literature. In Aim 3, we will develop a statistical method to construct meta-annotations that can effectively summarize high dimensional annotation data without losing interpretability. In Aim 4, we will apply these methods to genetic studies of vascular complications and autoimmune diseases in African American populations, with PubMed literature and various annotation datasets. The proposed research is innovative because it proposes a novel statistical framework that integrates multiple GWAS, biomedical literature, and annotation datasets to improve identification and interpretation of risk variants. The proposed research is significant because it is expected to help improve diagnosis and treatment of diseases with more effective identification of risk variants and enhanced understanding of common etiology among diseases.