Advancing Drug Repositioning for Alzheimer’s Disease using Real-world Data

项目总结： 阿尔茨海默病(AD)和AD相关痴呆(ADRD)是影响约 570万美国人。一般来说，五分之一的女性和十分之一的男性预计会患上AD/ADRD； 预计在未来20年内，患有AD/ADRD的人数将增长到1400万人。这个 AD/ADRD患者的生活质量逐渐下降，对AD/ADRD患者的护理强加于 给家庭照顾者、社区和医疗保健系统带来巨大的情感和经济负担。 然而，到目前为止，AD/ADRD患者还没有治愈的方法，甚至没有有效的治疗方法，这可能是由于 AD/ADRD发病机制复杂。随着药物开发成为 日益昂贵和耗时(估计成本从6.48亿美元8到25亿美元9和 新药的平均使用年限为9-12年)，旨在发现现有药物的新用途的药物再利用就是其中之一 加速AD/ADRD药物开发的潜在解决方案。然而，之前对毒品的尝试 根据组学数据重新调整AD/ADRD的用途到目前为止还没有成功，这表明动物模型 可能不会像希望的那样容易地转化为人类。可以加快药物开发的新方法 需要AD/ADRD。 在这项研究中，我们建议使用4种独特的EHR来检测可能被重新用于AD/ADRD的药物 数据集。这项研究将解决基于EHR的药物再利用的关键挑战，包括不完整 患者的信息和错误分类错误相关的偏差。目标1将专注于药物用途的重新调整 AD/ADRD知识库，用于从临床中提取危险因素的自然语言处理方法 叙述和表型算法，以准确识别MCI和AD/ADRD患者以支持患者 队列建设。在目标2中，我们将开发药物再利用方法，以解决高维 风险因素和错误分类错误相关的偏差，并将其应用于检测药物再利用信号 使用来自(1)One佛罗里达网络(2)Cerner Health Fact数据库的大量EHR，(3) EHR来自德克萨斯大学休斯顿健康科学中心的医生实践，(4)EHR数据来自 宾夕法尼亚大学。在目标3中，我们建议通过前瞻性的方法来验证排名靠前的信号 队列研究。我们将招募患者，并定期收集详细的实用信息和基因类型，以 验证识别出的药物信号的有效性。我们的研究的成功将：(1)产生一个知识库 及时更新AD/ADRD的危险因素、生物标志物、基因类型和药物信号，(2)开发一种开放的- 来源药物再利用包装-Rraider(利用电子健康重新利用影响阿尔茨海默病的药物 记录)用于AD/ADRD，以及(3)产生在前瞻性队列研究中验证的药物再利用信号， 这将为今后AD/ADRD的大规模国家试验的设计提供信息。

Project Summary: Alzheimer’s disease (AD) and AD-related dementias (ADRD) is the 6th leading cause of death affecting about 5.7 million Americans. Generally, one in five women and one in ten men are expected to develop AD/ADRD; and the number of people living with AD/ADRD is expected to grow to 14 million in the next two decades. The quality of life of AD/ADRD patients is gradually diminished and caring for AD/ADRD patients imposes tremendous emotional and financial burden on family caregivers, communities, and healthcare systems. However, up until now, there is no cure and not even effective treatment for AD/ADRD patients, probably due to the complex mechanisms involved in the pathogenesis of AD/ADRD. As drug development is becoming increasingly expensive and time-consuming (with estimated cost from $648 million8 to $2.5 billion9 and an average of 9-12 years for new drugs), drug repurposing, aiming to discover new uses of existing drugs, is one potential solution to speed up the drug development for AD/ADRD. However, previous attempts on drug repurposing for AD/ADRD based on omics data have not been successful so far, indicating that animal models may not translate to humans as readily as hoped. New methods that can speed up drug development for AD/ADRD are needed. In this study, we propose to detect drugs that can be potentially repurposed for AD/ADRD using 4 unique EHR data sets. This study will address the critical challenges of EHR-based drug repurposing including incomplete patient’s information and misclassification error associated bias. Aim 1 will focus on a drug repurposing knowledgebase for AD/ADRD, natural language processing methods to extract risk factors from clinical narratives, and phenotyping algorithms to accurately identify MCI and AD/ADRD patients to support the patient cohort construction. In Aim 2, we will develop drug repurposing methods that account for the high-dimensional of risk factors and misclassification error associated bias and apply them to detect drug repurposing signals using large collections of EHRs from (1) the OneFlorida network (2) the Cerner Health Facts database, (3) EHR from physician practice at University of Texas Health Science Center at Houston, and (4) EHR data from the University of Pennsylvania. In Aim 3, we propose to validate the top-ranked signals through a prospective cohort study. We will recruit patients and routinely collect detailed pragmatic information and genotypes to validate the efficacy of the identified drug signals. The success of our study will: (1) produce a knowledgebase with timely updated risk factors, biomarkers, genotypes, and drug signals for AD/ADRD, (2) develop an open- source drug repurposing package - RAIDER (Repurposing Alzheimer Impacting Drugs using Electronic health Records) for AD/ADRD, and (3) generate drug repurposing signals validated in a prospective cohort study, which will inform the design of future large-scale national trials for AD/ADRD.