Systems Pharmacology Modelling for Translating Animal Models of Neuroinflammation

慢性神经炎已经成为导致衰老和几种神经退行性疾病的主要因素，这些疾病的治疗选择仍然有限或根本不存在，包括阿尔茨海默病和多发性硬化症。虽然已经确定了一些遗传和外部风险因素，但确定的原因和治疗方法仍然难以捉摸。实验动物模型是检验疾病发生/发展假说、产生预防和干预的潜在靶点以及评估新化合物、现有药物(即改变用途)和创新药物组合的浓度-效应关系的关键工具。然而，很明显，没有一个动物模型可以完全概括与人类慢性炎症和神经变性相关的病理生理学、神经认知症状、药物反应的异质性以及浓度-效应关系。我们的目标是开发一种混合建模策略来评估可用于慢性神经炎症的各种动物模型，确定实验和计算平台的组合以最好地将临床前试验转化为人类，并使用最终模型来评估当前药物开发的靶点。我们的混合建模策略将结合定量系统药理学(QSP)建模和机器学习(ML)算法，将动物疾病进展和药物反应的分子决定因素与人类临床结果联系起来。这是一种新兴的范式，通过利用已知生物物理过程(QSP)的建模和自上而下的ML方法来弥合知识差距并生成新的可测试假设，在其他治疗领域显示出了希望。我们希望QSP/ML平台本质上是模块化的，这样它就可以被校准到多种神经退行性疾病。虽然有几个QSP模型已经被开发用于神经科学研发[6]，但它们没有考虑到神经炎成分，这些成分可能是许多以前有希望用于治疗多种神经退行性疾病的化合物失败的原因。我们已经展示了如何以战略性的方式使用多种建模方法，例如布尔网络分析和常微分方程式，将药物暴露和分子水平的事件联系起来，以探索治疗结果的异质性、替代剂量策略以及肿瘤学中的联合用药方案。最近，Mager博士(我们该项目的合作者)将他的QSP和基于网络的建模方法转变为神经科学，并开发了一个网络模型来识别药物靶点和重新定位的药物，以预防/治疗化疗引起的周围神经病。在这里，我们将继续这种整合多水平数据(包括转录和蛋白质水平)的方法，以控制免疫和神经系统之间复杂的相互作用因素，以响应神经退行性疾病和药物治疗。学生将负责开发一个QSP/ML平台，以研究慢性神经炎动物模型的翻译潜力。实验、文献和在线数据将被用来校准与慢性神经炎的发展有关的主要细胞信号通路的大型动态模型。这一模型反过来将提供更丰富的疾病生物学图景，并为药物的开发提供信息，这些药物在阿尔茨海默病、多发性硬化症、帕金森病和肌萎缩侧索硬化症中提供更大的治疗益处，减少安全担忧，并验证当前的治疗目标。我们预计该模型将用于神经科学部，以确定疾病进展的生物标记物，临床试验中的患者选择，以及AZ投资组合中不同项目的治疗反应。亚利桑那州神经科学系的科学家将提供神经炎症方面的生物学知识，建模专业知识，以及访问内部数据库。

Chronic neuroinflammation has emerged as a major contributing factor to aging and several neuro-degenerative diseases for which treatment options remain limited or non-existent, including Alzheimer's disease and multiple sclerosis. Although some genetic and extrinsic risk factors have been identified, definitive causes and cures remain elusive. Experimental animal models are a critical tool for testing hypotheses of disease initiation/progression, generating potential targets for prevention and intervention, and evaluating the concentration-effect relationships for new compounds, existing drugs (i.e., repurposing), and innovative drug combinations. However, it is clear that no one animal model can completely recapitulate the pathophysiology, neuro-cognitive symptoms, heterogeneity in drug responses, and concentration-effect relationships associated with chronic inflammation and neurodegeneration in humans. Our goals are to develop a hybrid modelling strategy to evaluate the various animal models available for chronic neuro-inflammation, identify a combination of experimental and computational platforms to best translate preclinical testing to humans, and to use the final models to evaluate current targets for drug development. Our hybrid modelling strategy will combine quantitative systems pharmacology (QSP) modelling with machine learning (ML) algorithms to bridge the molecular determinants of disease progression and drug response in animals to human clinical outcomes. This is an emerging paradigm that has shown promise in other therapeutic areas by leveraging the modeling of known biophysical processes (QSP) and top-down ML approaches for bridging knowledge gaps and generating new testable hypotheses. We intend for the QSP/ML platform to be modular in nature such that it could be calibrated to multiple neurodegenerative disorders. Although there are several QSP models that have been developed for neuroscience R&D [6], they do not account for the neuroinflammatory components that may underly the failure of many formerly promising compounds for multiple neurodegenerative disorders. We have shown how multiple modelling approaches, such as Boolean network analysis and ordinary differential equations, can be used in a strategic manner to link drug exposure and molecular level events to explore heterogeneity in treatment outcomes, alternate dosing strategies, and combination drug regimens in oncology. Recently, Dr Mager (our collaborator on this project) has transitioned his QSP and network-based modelling approaches to neuroscience and developed a network model to identify drug targets and a re-purposed drug to prevent/treat chemotherapy-induced peripheral neuropathy. Here, we will continue this approach of integrating multilevel data (including transcriptomics and protein levels) for factors governing the complex interplay between the immune and nervous systems in response to neurodegenerative disease and drug therapies.The student will be charged with developing a QSP/ML platform to study the translational potential of animal models of chronic neuroinflammation. Experimental, literature, and online data will be used to calibrate a large dynamic model of major cell signaling pathways implicated in development of chronic neuroinflammation. This model will in turn offer a richer picture of disease biology and inform the development of medicines that deliver greater therapeutic benefit with fewer safety concerns in Alzheimer's disease, multiple sclerosis, Parkinson's disease, and ALS, and the validation of current targets for therapy. We expect the model to be used in the neuroscience department for identifying biomarkers of disease progression, patient selection in clinical trials and response to treatment for the different project in AZ portfolio. Scientists in the neuroscience department at AZ will provide biological knowledge in neuroinflammation, modelling expertise, and access to internal databases.