An In Silico, Medical Record-based Model for Understanding the INitiation of Autoimmune Events (IMMUNE)

摘要 超过80%的接受免疫检查点抑制剂（ICI）治疗的癌症患者会出现免疫反应。 相关不良事件，包括治疗后的自身免疫。自身免疫性疾病的病因 对人类的了解很少，有效的治疗方法有限。近亲繁殖的老鼠是一个有价值的工具， 了解疾病的基本生物学机制，但对于了解人类 自身免疫因此，开发微创人体模型以提供对真实的世界的了解是至关重要的 自身免疫的机制、发展和治疗。电子健康的广泛应用 医疗保健中的电子病历（EHR）和为癌症患者收集的数据的深度，提出了一个重要的 有机会确定免疫治疗后发生自身免疫性疾病的风险因素。我们 该项目汇集了免疫学家，肿瘤学家，信息学家和机器学习专家团队 在EHR网络中工作，以确定接受ICI治疗的癌症患者队列。从 我们将设计和实施一个广泛而深入的EHR数据纵向数据库，包括 治疗和反应数据以及实验室结果，以便能够开发表型特征， 人类自身免疫性疾病发展的模型。本项目提案的首要目标是测试 使用与机器学习相结合的混合计算机/体内模型系统的可行性和有效性 作为了解自身免疫性疾病病因学的平台。在R61阶段，我们建议 鉴定在存在或不存在癌症的情况下发生类风湿性关节炎（RA）的患者， 队列，使用医生验证的癌症患者队列和EHR数据，并使用机器学习 在存在或不存在癌症和ICI治疗的情况下开发RA表型谱的策略。在 R33阶段，我们将开发和评估全球生物标志物耐受性破坏的表型特征， 机器学习并确定家族史是否是ICI后自身免疫发展的预测因素 疗法我们的建议是开发一种基于计算机的模型来探索自身免疫的发病，这是一个飞跃。 翻译免疫学与人类自身免疫性疾病机制的研究进展 通过利用EHR中收集的信息的力量来预测结果。表型 开发的配置文件可以显着加速采用ICI的精确医学方法， 潜在的自身免疫性疾病的基础上，个人的遗传，环境和社会信息。

Abstract More than 80% of cancer patients who undergo immune checkpoint inhibitor (ICI) therapy experience immune related adverse events, including autoimmunity, following treatment. The etiology of autoimmune disease in humans is poorly understood and effective treatments are limited. Inbred mice are a valuable tool for understanding basic biological mechanisms of disease, but are less effective for understanding human autoimmunity. Therefore, it is critical to develop minimally invasive human models to provide real world insights into the mechanisms, development and therapy for autoimmunity. The widespread use of electronic health records (EHRs) in healthcare and the depth of data collected for cancer patients, presents an important opportunity to identify risk factors for the development of autoimmune disease following immunotherapy. Our project brings together a team of immunologists, oncologists, informaticists and machine learning experts working within an EHR network, to identify a cohort of cancer patients who have undergone ICI therapy. From this cohort we will design and implement a broad and deep longitudinal database of EHR data, including treatment and response data and laboratory results, to enable the development of phenotypic profiles and models for autoimmune disease development in humans. The overarching goal of this project proposal is to test the feasibility and effectiveness of using a hybrid in silico / in vivo model system combined with machine learning strategies as a platform for understanding the etiology of autoimmune disease. In the R61 Phase we propose to identify patients who develop rheumatoid arthritis (RA) in the in the presence or absence of cancer, and control cohort, using a physician-validated cohort of cancer patients and data from EHR, and use machine learning strategies to develop phenotypic profiles for RA in the presence or absence of cancer and ICI therapy. In the R33 Phase we will and develop and assess phenotypic profiles for global biomarkers tolerance disruption using machine learning and determine if family history is a predictor of the development of autoimmunity following ICI therapy. Our proposal, to develop an in silico based model for exploring the onset of autoimmunity, makes a leap forward for translational immunology and the exploration of mechanisms of human autoimmune disease development by leveraging the power of the information collected in EHR to predict outcomes. The phenotypic profiles developed could significantly accelerate precision medicine approaches for employing ICIs that minimize the potential autoimmune disease based on personal genetic, environmental and social information.