Investigating a molecular basis for Alzheimer's disease subtypes using multiomic data integration and machine-learning

项目摘要 用多组学数据整合研究AD亚型的分子基础 和机器学习 尽管对临床前阿尔茨海默病(AD)疾病模型进行了密集的调查，但所有潜在的疾病- 修改药物在临床试验中失败了。大量的遗传学研究已经提出了许多生物学上的 然而，对于阿尔茨海默病的遗传病因，目前还没有达成共识。这很可能是因为 普遍认为阿尔茨海默病是一种单一的疾病，过于简单化，没有考虑到异质性致病因素 AD遗传结构的变异。高通量研究表明，AD是复杂的、非线性的结果 基因组、转录组、表观基因组和蛋白质组内部和之间的相互作用。虽然全基因组 协会研究已经成功地发现了与AD相关的基因，这些基因解释了 患者人口比例很小，“缺失遗传性”的问题仍然存在。因此，在目标1中，我 建议在具有机器学习的集成多组学框架中使用线性和非线性方法 确定AD中重要的途径。虽然几乎所有的AD患者都表现出标志性的b-淀粉样蛋白和 神经原纤维缠绕病理，它们也表现出显著的认知症状、行为和 神经生理学。鉴于此，我假设AD相关和免疫途径的个体间变异 在患者群体中驱动不同的疾病病因，最终导致共同的病理生理学。一 异质性的来源可能是在免疫途径不同地调节神经炎性反应 广告。在目标2中，我建议使用一种非监督分类方法来确定AD的亚型，基于 患者在基因组水平、成像数据和表型数据的路径变化上的相似性。具体来说，我 假设先天免疫途径中的致病变异在驱动不同的 患者之间的疾病病因。在目标3中，我建议通过产生蛋白质来表征每个基因组亚型 用于确定药物靶标优先顺序的互动网络。来自这些目标的知识将为当前AD的转变提供信息 通过提供针对AD特定基因组亚型的精确医学方法的药物开发范式 而不是“一刀切”的方法，这种方法迄今未能奏效。通过研究阿尔茨海默病的基因组异质性 这些目标有可能影响对有症状的AD患者的检测，并揭示更多 对AD复杂遗传结构的洞察。

PROJECT ABSTRACT Investigating a molecular basis for AD subtypes using multiomic data integration and machine-learning Despite intense investigation into preclinical Alzheimer’s Disease (AD) disease models, all potential disease- modifying drugs have failed in clinical trials. Numerous genetic studies have proposed a number of biological mechanisms, however there has been no consensus on the genetic etiology of AD. This is likely because the prevailing view of AD as a singular disease is oversimplified and does not consider heterogeneous pathogenic variation in AD genetic architecture. High-throughput studies indicate that AD is a result of complex, nonlinear interactions within and between the genome, transcriptome, epigenome, and proteome. While genome-wide association studies have successfully revealed genes associated with AD, these genes explain disease in a small proportion of the patient population, and the question of “missing heritability” remains. Thus, in Aim 1, I propose using linear and nonlinear methods in an integrated multiomics framework with machine learning to identify pathways significant in AD. While almost all AD patients present the hallmark b-amyloid and neurofibrillary tangle pathology, they also present significant variability in cognitive symptoms, behaviors, and neurophysiology. Given this, I hypothesize that inter-individual variation in AD-associated and immune pathways drives different disease etiologies across the patient population culminating in a common pathophysiology. One source of heterogeneity may be in immune pathways differentially regulating neuroinflammatory response during AD. In Aim 2, I propose using an unsupervised classification approach to determine subtypes of AD based on patient similarity in pathway variation across omic levels, imaging data, and phenotypic data. Specifically, I hypothesize that pathogenic variation within innate immunity pathways plays a critical role in driving different disease etiologies between patients. In aim 3, I propose characterizing each omic subtype by generating protein interaction networks for drug target prioritization. Knowledge from these aims will inform a shift in the current AD drug development paradigm by informing a precision medicine approach to target specific omic subtypes of AD instead of a “one size fits all” approach that has failed to date. Investigating genomic heterogeneity in AD through these aims has the potential to impact detection of pre-symptomatic AD individuals as well as reveal more insights into the complex genetic architecture of AD.