Big data screening for associations between medication use and ALS

项目摘要/摘要 我们的长期目标是确定可能影响发病率的药物或药物组合 肌萎缩侧索硬化症(ALS)或改变ALS的进展。在体外努力探索单身 基于已提出的治疗ALS的生物学机制的化合物在ALS治疗中的有效性是一种 有价值和有保证的方法，但它本身就很慢，因为需要一次一个地测试化合物 时间到了。它也没有探索组合的影响，也不能测试这些化合物在ALS中的作用 发病率。考虑到老年人服用的药物种类繁多，而且有可能不同 药物的组合可能是相关的，我们建议一个有价值的平行方法将是 流行病学筛选过程，以测试任何当前使用的药物是否与ALS的发病率有关 或者生存。这将类似于体外高通量筛选，但使用新的统计方法来 探索高维“大”流行病学数据(许多人、许多药物)，以了解与 肌萎缩侧索硬化症和肌萎缩侧索硬化症的存活率：具体地说，布尔逻辑回归和随机森林嵌套病例对照和 生存分析框架。这些方法允许高效地探索高维的、可能的 关于个别药物和不同药物组合之间的关联的相关数据 一个结果，在这里是ALS。为了实现这一点，我们建议使用两个并行的非常大的数据集 前瞻性和客观地收集药品和健康数据：丹麦注册系统和 以色列的Clalit卫生系统，总共有大约4,300例肌萎缩侧索硬化症病例和300,000多名控制者。通过 使用不同人群中的两个数据集，我们将增加识别因果关系的概率 通过鉴定那些在两个群体中都呈阳性的化合物。这项工作的结果是 确定当前使用的药物或这些药物的组合是否可能影响ALS和 与肌萎缩侧索硬化症共存。任何积极的结果都可能开辟新的研究途径，包括有针对性的临床试验 以及潜在的流行病学研究和实验室研究的新方向 机械装置。

PROJECT SUMMARY/ABSTRACT Our long-term goal is to identify medications, or combinations of medications, that may affect incidence of amyotrophic lateral sclerosis (ALS) or alter the progression of ALS. In vitro efforts to exploring single compounds for efficacy in ALS therapeutics based on proposed biological mechanisms for the disease is a valuable and warranted approach, but it is inherently slow because of the need to test compounds one at a time. It also does not explore effects of combinations and it cannot test the role of these compounds in ALS incidence. Given the wide array of medications that older adults take, and the possibility that different combinations of medications may be relevant, we propose that a valuable parallel approach would be an epidemiological screening process to test whether any currently used medications are related to ALS incidence or survival. This would be akin to in vitro high throughput screening, but using novel statistical approaches to explore high dimensional “big” epidemiological data (many people, many medications) for associations with ALS and ALS survival: specifically, boolean logic regression and random forests in a nested case-control and survival analysis framework. These approaches allow for efficiently exploring high-dimensional, likely correlated, data for associations between individual medications and different combinations of medications and an outcome, here ALS. In order to accomplish this, we propose to use two parallel very large data sets with prospectively and objectively collected pharmaceutical and health data: The Danish Registry System and the Clalit Health System in Israel, with a total of approximately 4,300 ALS cases and over 300,000 controls. By using the two data sets in different populations we will increase the probability of identifying causally related compounds by identifying those that screen positive in both populations. The results of this work have the possibility to identify currently used medications or combinations of these medications that can affect ALS and survival with ALS. Any positive results could open up new research avenues that include targeted clinical trials as well as potentially new directions for epidemiological studies and laboratory research into underlying mechanisms.