Multi-Cohort, Network-Guided Regression for GE/GG Interactions in Disease Traits

描述（由申请人提供）：复杂疾病被认为是由遗传和环境性质以及生活方式等多种因素引起的。例如心血管疾病（如心脏病、高血压）、代谢疾病（如糖尿病）、神经疾病（如阿尔茨海默病、帕金森病和自闭症）和癌症。这些疾病给家庭和社会造成的经济损失是惊人的，而对个人及其亲人造成的身体和精神损失是无法估量的。虽然在早期，人们预计解码人类基因组将是对复杂疾病的原因获得重大见解的直接先驱，但现在很明显，这一领域的进展必须来自通过生物途径系统和相关网络揭示基因与环境的相互作用，以及基因之间的相互作用。我们的目标是开发一种新的网络引导统计方法，以促进发现与人类疾病相关的复杂数量性状相关的基因-环境（GxE）和基因-基因（GxG）相互作用。具体而言，我们将(1)开发一类稀疏的、网络引导的回归模型，用于检测GxE和GxG相互作用；(2)通过开发两阶段荟萃分析策略，将该回归框架的适用性扩展到多个队列；(3)在模拟中评估总体方法，并使用来自两个特定疾病领域的数据：糖尿病相关的数量特征和肺部数量特征和疾病。数据分析将与弗雷明汉心脏研究中心的同事以及MAGIC和CHARGE两个协会共同完成。成功完成拟议的研究将产生一套高度新颖和连贯的工具（包括软件实施），用于在当前大规模、多队列关联分析中检测与人类疾病相关的GxE和GxG相互作用的原则性和生物学知情的两阶段方法。最终，我们的工作应该通过其对整个过程早期阶段的根本性影响，帮助显著加快靶向治疗和个性化医疗策略的发展。

DESCRIPTION (provided by applicant): Complex diseases are believed to be caused by a number of factors of both genetic and environmental nature, as well as lifestyle. Examples include cardiovascular (e.g., heart disease, hypertension), metabolic (e.g., diabetes), and neurological (e.g., Alzheimer's, Parkinson's, and autism) diseases, and cancer. The financial costs of these diseases on families and society are staggering, while their physical and emotional toll on individuals and their loved ones is incalculable. While early on it was expected that decoding the human genome would be an immediate precursor to gaining significant insight into the causes of complex diseases, it is clear now that progress in this area must come from unraveling the interplay of genes with environment, as well as with each other, through the system of biological pathways and related networks. Our goal in this proposal is the development of a novel network-guided statistical methodology to facilitate the discovery of gene-environment (GxE) and gene-gene (GxG) interactions associated with complex quantitative traits associated with human disease. Specifically, we will (1) develop a class of sparse, network-guided regression models for detection of GxE and GxG interactions, (2) extend the applicability of this regression framework to multiple cohorts through the development of a two-stage meta-analysis strategy, and (3) assess the overall methodology both in simulation and using data from two specific disease areas: diabetes-related quantitative traits and pulmonary quantitative traits and diseases. The data analyses will be done in conjunction with colleagues at the Framingham Heart Study and two consortia: MAGIC and CHARGE. Successful completion of the proposed research will yield a highly novel and coherent set of tools (including software implementation) for a principled and biologically- informed two-stage approach to detecting GxE and GxG interactions associated with human disease in current large-scale, multi-cohort association analyses. Ultimately, our work should help to significantly accelerate the development of targeted therapies and personalized medicine strategies, through its fundamental impact on the early stages of the overall process.