Harnessing Diverse BioInformatic Approaches to Repurpose Drugs for Alzheimers Disease

摘要 阿尔茨海默病患者脑部基因组、转录组和蛋白质组的研究 (Ad)通过强大的计算工具有可能发展新的知识，包括 确定可能参与疾病启动和/或发展的途径和靶点。 挑战是找到影响这些通路的药物，然后验证这些通路的重要性 -区分主要疾病驱动因素和次要事件。改变FDA批准的药物的用途是其中之一 在概念验证和最终治疗的临床试验中探索潜在途径的方法。在这里，我们 建议通过三个集成的、 互补的信息学方法。具体地说，我们将应用经典和网络感知(预先加载) 机器学习方法识别阿尔茨海默病不同阶段大脑中改变的通路和靶点 使用通过Synapse(AIM)提供的Accelerating Medicines Partnership-AD数据进行的疾病进展 1)；我们将使用系统药理学方法来发现先导化合物在 使用无偏的RNA-SEQ、蛋白质组学和成像研究的人类神经元和神经胶质细胞类型 通径分析(目标2)。这两个目标都有两种方法：数据驱动、假设生成 辨别与疾病相关的药物信号的分析；以及假设检验，其中一个人的积极发现 使用其他方法对方法进行评估，以评估严密性和重复性。此外，rna-seq 以及在暴露于潜在致病因素后培养的人CNS细胞类型中收集的蛋白质组学数据 和/或FDA批准的AIM 2中的药物将作为CNS细胞类型派生的前体反馈到AIM 1中，以提炼 预测性模型。在目标3中，我们将开发新的信息学策略，在电子病历中进行电子药物试验。 具有基于组学数据集和现有文献的验证假设的“预期”结果的数据， 使用两个大数据集：英国20年来CPRD对2000万国民医疗服务患者的纵向记录， 和RPDR数据库(基于合作伙伴医疗保健)，包含600万人，跟踪了20多年。这 综合信息学课程弥补了每种个别信息学方法的局限性 促进对已知和新AD途径的“先导化合物”的发现和批判性评估。执行，执行 在这一战略中，我们组建了一支多地点、多学科的团队，拥有从临床护理到临床护理的专业知识 计算机科学和系统药理学。一些团队成员是AD专家，另一些人则带来了 一个局外人的视角。最后，作为可交付内容，我们将创建开放源码的数据包来发布所有 支持证据、软件和数据的来源符合公平(可发现、可访问、 可互操作和可重现)标准，通过Synapse和MGH开发的阿尔茨DataLens平台实现 (目标4)。这些数据包将有助于确定后续临床和转化性研究的优先顺序，包括 与业界或整个社区成员的合作涉及到新的临床试验。

Abstract The exploration of genomes, transcriptomes, and proteomes derived from brains with Alzheimer's disease (AD) by powerful computational tools has the potential of developing new knowledge, including the identification of pathways and targets that may be involved in the initiation and/or progression of the disease. The challenge is to find drugs that impact those pathways, and then validate the importance of those pathways – distinguishing primary disease drivers from secondary events. Repurposing FDA-approved drugs is one approach to probe potential pathways in proof of concept, and ultimately therapeutic, clinical trials. Here, we propose to discover and validate hypotheses for drug repurposing in AD through three integrated, complementary informatics approaches. Specifically, we will apply classical and network aware (prior-loaded) machine learning approaches to identify pathways and targets altered in AD brains at different stages of disease progression using data from Accelerating Medicines Partnership-AD available through Synapse (Aim 1); we will use systems pharmacology approaches to discover the target selectivity of lead compounds in human neuronal and glial cell types using unbiased RNA-seq, proteomic and imaging studies followed by pathway analysis (Aim 2). Each of these two Aims has two approaches: data-driven, hypothesis-generating analyses to discern disease-relevant drug signals; and hypothesis-testing in which positive findings from one approach are evaluated using the other approaches to assess rigor and reproducibility. Moreover, RNA-seq and proteomic data collected in cultured human CNS cell types following exposure to potential disease drivers and/or FDA-approved drugs in Aim 2 will be fed back into Aim 1 as CNS-cell type-derived priors to refine the predictive models. In Aim 3, we will develop new informatics strategies to conduct in-silico drug trials in EHR data with “prospective” outcomes to validate hypotheses based on the omics data sets and extant literature, using two big data sets: the UK 20 year CPRD longitudinal records of 20M National Health Service patients, and the RPDR Database (based at Partners Healthcare) with 6 M individuals followed for over 20 years. This integrated informatics program compensates for the limitations of each individual informatics approach to promote discovery and critical evaluation of “lead compounds” for known and novel AD pathways. To execute this strategy, we have assembled a multi-site, multi-disciplinary team with expertise ranging from clinical care to computer science and systems pharmacology. Some of the team members are AD experts and others bring an outsider's perspective. Finally, as a deliverable, we will create open-source data packages to release all the supporting evidence, software, and data with provenance in accordance with FAIR (findable, accessible, interoperable and reproducible) standards through Synapse and the AlzDataLens platform developed at MGH (Aim 4). These data packages will help to prioritize follow on clinical and translational studies including collaborations with industry or members of the community at large involved in new clinical trials.