CAREER: A systems engineering approach to elucidate and treat multi-factorial pathology

全世界有数百万患者患有多因素疾病，这是一种没有单一原因而是有许多促成因素的疾病。 由于其复杂性，大多数多因素疾病目前无法治愈。例子包括阿尔茨海默氏痴呆症（AD）、额颞叶痴呆症（FTD）和肌萎缩侧索硬化症（ALS）的破坏性神经退行性疾病，这些疾病都影响大脑功能和执行正常日常任务的能力。在传统的实验室或临床环境中，有效地测量整个疾病过程中的多个同时起作用的因素极具挑战性。CAREER项目的目标是开发新的复杂计算机模型，整合并同时分析数千项研究的数据，这些研究检查了在实验室或诊所测量的单个疾病因素。开发的计算机模型优先考虑最有希望的因素，并制定最佳的组合治疗策略。计算机优先级提高了临床试验成功的可能性，并加快了新的治疗方法对患者的可用性。虽然该项目的重点是预测AD，FTD和ALS的治疗方法，但开发的新技术可以应用于许多其他多因素疾病。该项目的教育活动侧重于本科生研究实习课程，以增加机会;残疾学生的专业指导;通过地方和国家组织对AD，FTD和ALS患者的综合宣传;通过开发新的综合类，将治疗设计与医学院临床神经疾病讲座相结合，改善神经工程的大学教育。除研究生研究助理外，这一项目估计还将为约100名本科生和高中实习生提供STEM研究实习机会，重点关注代表性不足的群体，研究者的长期目标是利用“病理动力学”（病理生理学的一个分支，研究运动，平衡，或生理系统在病理力作用下的体内平衡）来增强预测医学，其主要目的是改善，通过开发预测疾病进展和治疗反应的计算机模型，加快和个性化医疗保健。 为了实现这一目标，这个CAREER项目的目标是构建新的文献挖掘和预测医学模型，利用病理动态来解决多因素疾病。 大多数多因素疾病是难治性的，对传统的治疗方法没有反应。 该项目的驱动假设是，病理动力学是解锁独特特征的关键，这些特征可以区分一系列具有相似症状，病因和生物标志物的多因素疾病，但由于缺乏敏感和特异性的临床诊断测试，目前是临床“排除诊断”。 三个多因素神经病理学测试案例-阿尔茨海默病（AD），肌萎缩性侧索硬化症（ALS）和额颞叶痴呆症（FTD）-将用于表征基于病理动力学的模型改善诊断，预后和治疗预测的能力。 研究计划有三个目标。 第一个目标是建立数据库来捕获、量化和聚合整个领域的文献。 研究者的优化的学生驱动的装配线（高中和本科生的生物定位任务培训的层次结构）将完全重新捕获期刊文章数据沿着的关键实验方法/协议，使有意义的数据汇总和分析。 组装线将首先完成完整的已发表临床前数据重新获取和ALS、AD和FTD的相应数据库（约40，000篇文章），然后构建综合临床数据库，其中包括AD、ALS和FTD的去识别患者数据。 还将努力开发和整合额外的生物制药自动化，以实现全面数据回收，目标是将生物制药自动化从40%提高到75%。 第二个目标是开发文献关系提取和排名的协议。使用联合医学语言系统本体进行关键字分类，并采用适应性无监督等级聚合对感兴趣的关系进行优先级排序，使用语义推理网络进行文本挖掘，以识别来自3000多万篇PubMed文章的多标量关系。 第三个目标是使用目标1中管理的数据和目标2中确定和排名的关系构建多因素疾病的“病理动力学”模型。 将为病理动力学表型构建无监督模型，并为诊断和治疗预测构建监督机器学习模型。 这些模型将用于比较文献关系与实验测量关系的排名。 总之，该项目的成果包括：用于AD、ALS和FTD语料库全面管理的新型多标量数据库;新的生物定位自动化和基于关系的文献挖掘技术;以及AD、ALS和FTD的基于从头系统动力学的临床前和临床预测医学模型，其可用于预测病因、诊断、治疗，该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Millions of patients suffer worldwide from multi-factorial disease, which is a disease with no single cause but rather numerous contributing factors. Due to their complex nature, most multi-factorial diseases are currently incurable. Examples include the devastating neurodegenerative diseases of Alzheimer’s Dementia (AD), frontotemporal dementia (FTD), and Amyotrophic Lateral Sclerosis (ALS), which all impact brain function and the ability to perform normal daily tasks. Effectively measuring multiple simultaneous contributing factors throughout the disease course is extremely challenging in a traditional lab or clinical setting. The goal of this CAREER project is to develop new complex computer models that integrate and simultaneously analyze data from thousands of studies examining individual disease factors measured in the lab or clinic. The developed computer models prioritize the most promising factors and develop optimal combination treatment strategies. Computer prioritization increases the likelihood of clinical trial success and expedites the rate of new treatment availability to patients. While this project focuses on predicting treatments for AD, FTD, and ALS, the developed new technology can be applied to numerous other multi-factorial diseases. Educational activities for this project focus on undergraduate research internship curricula to increase opportunities; professional mentoring of students with disabilities; integrated advocacy of patients with AD, FTD, and ALS through local and national organizations; and improved collegiate education for neuroengineering via development of a new integrative class that integrates therapy design with medical school lectures on clinical neurologic disease. In addition to graduate research assistants, this project is estimated to provide STEM research internships for about 100 undergraduates and high school interns with an emphasis on under-represented groups.The investigator’s long-term goal is to use “pathology dynamics” (a branch of pathophysiology that deals with the motion, equilibrium, or homeostasis of physiological systems under the action of pathological forces) to enhance predictive medicine, whose primary purpose is to improve, expedite and personalize healthcare by developing computer models that forecast disease progression and treatment response. Towards this goal, the goal of this CAREER project is to construct new literature mining and predictive medicine models that leverage pathology dynamics to tackle multi-factorial disease(s). Most multi-factorial diseases are intractable and not responsive to traditional therapeutic approaches. The project’s driving hypothesis is that pathology dynamics is the key to unlocking unique signatures that can differentiate a spectrum of multi-factorial diseases that share similar symptoms, etiology, and biomarkers, but are currently clinical “diagnoses of exclusion” due to the lack of sensitive and specific clinical diagnostic tests. Three multi-factorial neuropathology test cases--Alzheimer’s Disease (AD), Amyotrophic Lateral Sclerosis (ALS), and frontotemporal dementia (FTD)--will be used to characterize the ability of pathology dynamics-based models to improve diagnostic, prognostic, and therapeutic prediction. The Research Plan is organized under three Aims. The FIRST Aim is to construct databases to capture, quantify, and aggregate entire fields’ literature. The investigator’s optimized student-driven assembly line (a hierarchy of high school and undergraduate students trained for biocuration tasks) will fully recapture journal article data along with key experimental methods/protocols, which enable meaningful data aggregation and analysis. The assembly line will first complete full published preclinical data recapture and corresponding databases for ALS, AD, and FTD (approximately 40,000 articles), which will be followed by construction of integrative clinical databases that consist of de-identified patient data for AD, ALS, and FTD. Efforts will also be made to develop and integrate additional biocuration automation for full data recapture with a goal of increasing biocuration automation from 40% to 75%. The SECOND Aim is to develop protocols for literature relationship extraction and ranking. Text mining with semantic inference networks will be used to identify multi-scalar relationships from 30+ million PubMed articles using the United Medical Language System ontology for keyword categorization and adapted unsupervised rank aggregation to prioritize relationships of interest. The THIRD Aim is to construct “pathology dynamics” models for the multi-factorial diseases using the data curated in Aim 1 and the relationships identified and ranked in Aim 2. Unsupervised models will be constructed for pathology dynamics phenotyping and supervised machine learning models will be constructed for diagnostic and therapeutic prediction. The models will be used in comparing rankings of literature relationships to experimentally measured relationships. In summary, the deliverables of this project include: novel multi-scalar databases for full curation of the AD, ALS, and FTD corpuses; new biocuration automation and relationship-based literature mining technology; and de novo systems-dynamics based preclinical and clinical predictive medicine models of AD, ALS, and FTD that can be used to predict etiology, diagnosis, treatment, and prognosis.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.