Systems Biology Modeling of Severe Community-Acquired Pneumonia

项目概要/摘要-项目1 继发于严重急性呼吸道感染的大流行性社区获得性肺炎（CAP） 冠状病毒2型综合征（SARS-CoV-2）的出现使CAP的公共卫生重要性成为焦点。 肺炎治疗成功临床应答（SUCCESSful Clinical Response in Pneumonia Therapy，CITMT）系统生物学中心 开发了一个强大的研究基础设施，前瞻性地收集了1，567个系列远端呼吸道样本， 595例需要机械通气的重度CAP和医院获得性肺炎（HAP）患者， 使用最先进的多组学方法分析这些临床样本。我们利用这些数据 生成SARS-CoV-2发病机制的系统模型，并将其应用于成功的 一种钙释放激活通道抑制剂，在一个实验组中， 第二阶段试验。在Super-CRT 2中，我们建议利用和扩展纵向临床和 分子数据。通过将机器学习应用于临床数据，我们观察到重度 肺炎在其病程中经历不同的、临床可识别的状态之间的转变。 住院治疗与或多或少有利的结果相关。这些过渡将作为 模型的基础，包括BAL和血清分析产生的初步数据，包括 超过50万支气管肺泡灌洗液细胞的单细胞RNA测序、细胞因子水平、蛋白质组学、T 细胞表观基因组和微生物组分析。我们将使用这些临床和分子数据来验证假设 应用于疾病发病机制的潜在空间模型的机器学习方法可以识别 临床治疗期间有利和不利临床转变/结局的分子预测因子 CAP的过程。一个必然的假设是，在对照临床试验中，这些决定因素的扰动 药物干预将允许模型预测能力的迭代。我们将解决这些问题 三个具体目标的假设： 目标1.确定有利和不利临床转变/结局的临床预测因素， 需要住院治疗的CAP患者的病程。 目标二。确定预测有利或不利的不同宿主或病原体基因组特征 严重CAP患者的临床转变/结局。 目标3.确定可用于现有或新开发的治疗的靶向途径 治疗学 CRT 2利用临床医生、生物学家和数据科学家之间的成功合作， 数据，处理远端肺样本，并将不同的数据集集成到潜在空间模型中，以开发大型 肺炎的规模模型，可以迅速转化为护理途径和新的疗法。

Project Summary/Abstract – Project 1 Pandemic community-acquired pneumonia (CAP) secondary to infection with the severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) brought the public health importance of CAP into sharp focus. Investigators in the Successful Clinical Response in Pneumonia Therapy (SCRIPT) systems biology center developed a robust research infrastructure to prospectively collect 1,567 serial distal respiratory samples from 595 patients with severe CAP and hospital acquired pneumonia (HAP) requiring mechanical ventilation and analyze these clinical samples using state-of-the art multi-omics approaches. We leveraged these data to generate a systems model of SARS-CoV-2 pathogenesis and applied it toward a successful clinical trial of Auxora, a calcium release activated channel inhibitor, that resulted in a 53% reduction in 30-day mortality in a phase II trial. In Super-SCRIPT (SCRIPT2), we propose to leverage and expand the longitudinal clinical and molecular data in SCRIPT. By applying machine learning to clinical data, we observe that patients with severe pneumonia undergo transitions between distinct, clinically recognizable states over the course of their hospitalization that are associated with more or less favorable outcomes. These transitions will serve as the foundation for a model incorporating preliminary data generated from BAL and serum analysis that includes single-cell RNA-sequencing of more than 500,000 bronchoalveolar lavage cells, cytokine levels, proteomic, T cell epigenomic, and microbiome analyses. We will use these clinical and molecular data to test the hypothesis that machine learning approaches applied to a latent space model of disease pathogenesis can identify molecular predictors of favorable and unfavorable clinical transitions/outcomes during the clinical course of CAP. A corollary hypothesis is that perturbations of these determinants during controlled clinical trials of pharmacologic interventions will allow iteration of the models’ predictive capabilities. We will address these hypotheses in three Specific Aims: Aim 1. To identify clinical predictors of favorable and unfavorable clinical transitions/outcomes over the course of CAP in patients requiring hospitalization. Aim 2. To determine distinct host or pathogen genomic features that predict favorable or unfavorable clinical transitions/outcomes in patients with severe CAP. Aim 3. To identify pathways that can be targeted for therapy with existing or newly developed therapeutics. SCRIPT2 draws on successful collaborations between clinicians, biologists and data scientists to organize clinical data, process distal lung samples and integrate disparate datasets into latent space models to develop large scale models of pneumonia that can be rapidly translated into care pathways and novel therapies.