TargetAD: A systems multi-omics approach to drug repositioning in Alzheimer's disease

项目摘要 迟发性阿尔茨海默病（AD）是一种进展缓慢、无法治疗的神经退行性疾病， 影响着当今老龄化人口的很大一部分。数百项临床试验和大量投资 到目前为止，药物开发工作还没有产生一种单一的疾病修饰疗法， 对疾病有很大的好处。药物重新定位，将已批准的药物应用于新的 疾病背景下，作为确定AD治疗方案的有希望的替代方案，已获得越来越多的关注。 为了在AD中成功地进行药物干预，药物或药物组合需要靶向复合物 在AD中以特定方式观察到的分子变化。为了确定发挥这些预期作用的药物， 详细了解生物系统下跨调控层的分子网络， 必需的.然而，这些网络并不容易获得，并且分散在数百项研究中， 复杂的数据库。为了应对这一挑战，我们提出了TargetAD，这是一个基于网络的框架， 这种分子网络来自遗传关联，共表达/相关网络，代谢途径， 基因调控数据、蛋白质-蛋白质相互作用以及组织特异性基因和蛋白质表达数据 与AD多组学关联以及药物-药物靶标数据和分子药物特征增强。 我们将通过利用内部产生的大规模、多组学关联结果的力量来实现这一目标。 NIH的大型"加速药物伙伴关系-阿尔茨海默病"倡议和其他大规模的 基于人口的研究。收集的证据将存储在一个可公开访问的图形数据库中， 然后，我们用于识别候选药物或药物组合（"候选物"）。 通过开发一种新的基于网络的机器学习方法，我们将对候选人进行排名， 数据库通过其概率以有益的方式影响AD网络。高级候选人将在 根据全面的优先顺序管道。为此，我们将回顾性调查是否 接受重新定位候选者的个体的纵向AD相关生物标志物特征显示了以下证据： 健康老龄化在AD的大型研究。这些分析将通过审查后， 死亡神经病理学负担支持候选物的有益作用。为了增加功率， 覆盖范围的候选人，我们将进一步分析电子健康记录从英国生物银行的额外 证据将与一个专家小组讨论，选出三名最有希望的候选人。这些 将通过AD动物模型的临床前验证研究进行评价。 总之，多学科专业知识的独特组合，对高知名度数据集的访问， 先进的计算集成管道将使我们能够识别AD中受到干扰的分子途径， 可被药物重新定位候选物靶向，因此是用于临床试验测试的主要候选物。

Project Summary Late-onset Alzheimer's Disease (AD) is a slowly progressing, untreatable neurodegenerative disorder that affects a substantial fraction of the aging population today. Hundreds of clinical trials and massive investments into drug development efforts have so far not resulted in a single disease-modifying therapy that showed a significant beneficial effect on the disease. Drug repositioning, the application of approved drugs in a novel disease context, has gained increasing attention as a promising alternative to identify treatment options for AD. For successful pharmaceutical intervention in AD, a drug or drug combination needs to target the complex molecular changes observed in AD in a specific manner. To identify drugs exerting these desired effects a detailed understanding of the molecular networks across regulatory layers that underly the biological system is required. However, these networks are not readily available and are scattered across hundreds of studies and complex databases. To address this challenge, we propose TargetAD, a network-based framework that builds this molecular network from genetic associations, co-expression/correlation networks, metabolic pathways, gene regulation data, protein-protein interactions, and tissue-specific gene and protein expression data augmented with AD multi-omics associations, as well as drug-drug target data and molecular drug signatures. We will achieve this by leveraging the power of large-scale, multi-omics association results generated within NIH's large “Accelerating Medicines Partnership - Alzheimer's Disease” initiative and other large-scale population-based studies. The collective evidence will be stored in a publicly accessible graph database, which we then use for the identification of candidate drugs or drug combinations (“candidates”). Through the development of a novel network-based machine-learning method, we will rank candidates in the database by their probability to affect AD networks in a beneficial way. High-ranking candidates will be subjected to a comprehensive prioritization pipeline. To this end, we will retrospectively investigate whether longitudinal AD-related biomarker profiles of individuals who took a repositioning candidate show evidence for healthier aging in large studies of AD. These analyses will be complemented by examining whether the post- mortem neuropathological burden supports a beneficial effect of the candidate. To increase power and coverage of candidates, we will further analyze electronic health records from the UK Biobank for additional evidence. The three most promising candidates will be selected in discussion with a panel of experts. These will be evaluated by preclinical validation studies in animal models of AD. In summary, the unique combination of multidisciplinary expertise, access to high-profile datasets and advanced computational integration pipelines will allow us to identify molecular pathways disturbed in AD that are targetable by drug repositioning candidates, which thus are prime candidates for testing in clinical trials.