SOFTWARE FOR LARGE-SCALE INFERENCE OF THE GENETICS OF LIFESTYLE MEASURES, BIOMARKERS, AND COMMON AND RARE DISEASES

世界各地大规模的种群生物库，疾病集中在NHGRI基因组测序计划 (GSP)和美国的我们所有人精确医学倡议项目将产生大量基因组 结合疾病结果和其他健康衡量标准的数据集。这些基因组研究将 确定与健康和疾病相关的基因组变异。然而，他们在所有背景下的联系 如果对数据进行单独分析，确定的可能关联仍不清楚。有一个不断增长的 认识到大多数性状都是多基因的。此外，人们越来越认识到，多效性是普遍存在的。 出于隐私方面的考虑，共享所有可能的基因和表型数据是一件具有挑战性的事情。方法可以 对汇总级数据进行推断，例如p值、效果大小估计和频率，将有助于我们 了解人类疾病和健康的遗传学。在这里，我们建议开发软件，用于 生活方式指标、生物标志物以及常见和罕见的遗传学的大规模推断 疾病。实现这一目标需要医学和人口遗传学、统计学方法方面的专业知识 开发，以及大型数据库管理方面的专业知识。该项目有三个主要目标。 首先，我们将创建全球生物库引擎：一个强大的、交互式的Web平台，用于推理 生活方式、生物标志物、常见病和罕见病的遗传学研究。我们将通过以下方式扩展功能 实施用于标记变体和表型的质量控制可视化和方法。我们将添加 研究设计工具，使用经验数据来估计统计能力，并创建灵活的框架 联合分析多种表型，同时控制假阳性和假阴性的统计模型 调查结果。其次，我们将提高全球生物库引擎的性能、可扩展性和可访问性 方便今后人群生物库和有针对性的常见病和罕见病。我们将创建一个托管的、 安全、经济高效的基于云的社区资源，并设计一个数据库系统，以减少 加载遗传关联研究的时间从几小时到几分钟，并允许传输统计数据 算法直接转化为遗传数据。最后，我们将改进基因组解释、可视化和数据 共享，通过实施新颖的分析大幅提高翻译发现的速度 方法：研究方法。我们将支持新的变体标注方法，并整合编码和非编码信息， 包括来自大规模表观基因组学研究的数据，用于变异和基因水平的推断。我们将实施 用概率编程语言实现的新的贝叶斯统计模型，稀疏规范 相关分析和截断奇异值分解。皮里瓦斯和他的团队有足够的 NIH资助的财团的经验，他们致力于NIH及其资助的整体使命 调查人员将发现新的知识，这些知识将为每个人带来更好的健康。

Large-scale population biobanks around the world, the disease focused NHGRI Genome Sequencing Program (GSP), and the United States’ All of Us Precision Medicine Initiative project will generate massive genomic datasets combined with disease outcomes, and other health measurements. These genomic studies will identify genomic variants relevant to health and disease. However, their association in the context of all possible associations identified will remain unclear if the data are separately analyzed. There is a growing recognition that most traits are polygenic. In addition, it is increasingly appreciated that pleiotropy is pervasive. Due to privacy concerns, it is challenging to share all possible genotype and phenotype data. Methods that can perform inference on summary level data, e.g. p-values, effect size estimates, and frequency, will facilitate our understanding of the genetics of human diseases and health. Here, we propose to develop software for large-scale inference of the genetics of lifestyle measures, biomarkers, and common and rare diseases. Achieving this goal requires expertise in medical and population genetics, statistical methods development, and expertise in management of large-scale databases. The project has three main objectives. First, we will create Global Biobank Engine: a powerful, interactive web platform for inference of the genetics of lifestyle measures, biomarkers, common and rare diseases. We will expand the features by implementing quality control visualizations and methods for flagging variants and phenotypes. We will add tools for study design that use empirical data to estimate statistical power, and create a flexible framework for statistical models that jointly analyze multiple phenotypes while controlling for false positive and negative findings. Secondly, we will improve Global Biobank Engine performance, scalability, and accessibility to facilitate future population biobanks and targeted common and rare disease. We will create a hosted, secure, and cost-effective cloud-based community resource, and design a database system that reduces the loading time for genetic association studies from hours to minutes and allows for streaming of statistical algorithms directly to genetic data. Lastly, we will improve genomic interpretation, visualization, and data sharing to dramatically increase the rate of translational discoveries by implementing novel analysis methods. We will support new variant annotation methods and integrate coding and non-coding information, including data from large-scale epigenomics studies, for variant and gene level inference. We will implement new Bayesian statistical models implemented in probabilistic programming languages, sparse canonical correlation analysis, and truncated singular value decomposition. PI Rivas and his team have ample experience with NIH-funded consortia, and they are dedicated to the overall mission of NIH and its funded investigators to uncover new knowledge that will lead to better health for everyone.