Imaging and multi-omics analyses to identify molecular subtypes of distinct emphysema patterns

项目总结/摘要 慢性阻塞性肺疾病（COPD）是一种进行性、使人衰弱的疾病，急需治疗- 修改治疗。肺气肿，COPD受试者中常见的进行性肺破坏， 预示着预后不佳该项目将利用两个大型的良好的表型，NHLBI资助的研究（ COPDGene and Lung Tissue Research Consortium（LTRC））和该团队在现代医学领域的广泛专业知识， 成像技术、多组学数据分析、机器学习方法和体外功能验证。 本申请的总体目标是鉴定新的多组学生物标志物和多组学基因的分子亚型。 利用系统生物学方法， 了解多种组学数据类型之间的关系。在目标1中，我们将应用局部直方图（LH） 胸部计算机断层扫描（CT）量化方法，以生成小叶中心的成像表型， 全小叶和间隔旁肺气肿。我们将对这些肺叶LH数据进行聚类， LH模式相似的受试者组。然后，我们将测试确定的单组学关联 肺气肿集群与遗传变异，DNA甲基化标记，端粒长度，基因表达， 外周血和肺组织样本中的蛋白质组学。目标2将开发和评估肺组织， 基于血液的多组学机器学习模型，用于肺气肿模式的可靠临床预测。及时 诊断需要基于血液的预测模型，因为它可以在CT扫描的受试者中识别肺气肿， 没有临床指征。这也将克服辐射暴露和假阳性结果的问题 与CT扫描有关。目标3将通过应用一种新的分子生物学方法发现肺气肿的亚型， 创新的、可解释的机器学习算法，捕获方向特征交互，并提供 网络表示的分子决定因素的肺气肿亚型。然后我们将执行群集 对双变量Shapley网络表示进行分析，以根据以下内容识别不同的受试者亚组： 图的相似性。为了确认已鉴定途径的关键调节因子，我们将进行靶基因的筛选。 在气道上皮细胞和肺成纤维细胞中的沉默和过表达研究。基因会 优先考虑利用现有生物学知识和网络分析进行功能验证。通过 结合创新，尖端的数据生成，分析方法和功能验证， 该项目将通过加强肺气肿表型和多组学分析， 更可靠的预测和更好地理解疾病病理学。这些知识将为我们 用于开发急需的新颖和个性化的治疗策略。

PROJECT SUMMARY/ABSTRACT Chronic obstructive pulmonary disease (COPD) is a progressive, debilitating disease in critical need of disease- modifying treatments. Emphysema, progressive lung destruction commonly encountered in subjects with COPD, portends a poor prognosis. This project will leverage two large well-phenotyped, NHLBI-funded studies (the COPDGene and Lung Tissue Research Consortium (LTRC)) and the team’s extensive expertise in modern imaging techniques, multi-omics data analysis, machine learning approaches, and in vitro functional validation. The overall objective of this application is to identify novel multi-omics biomarkers and molecular subtypes of centrilobular, panlobular, and paraseptal emphysema patterns utilizing a systems biology approach to understand relationships between the multiple omics data types. In Aim 1, we will apply the local histogram (LH) chest computed tomography (CT) quantification method to generate imaging phenotypes of centrilobular, panlobular, and paraseptal emphysema in each lung lobe. We will cluster these lobar LH data to identify distinct groups of subjects with similar LH patterns. We will then test for single-omics associations of the identified emphysema clusters with genetic variants, DNA methylation marks, telomere length, gene expression, and proteomics in peripheral blood and lung tissue samples. Aim 2 will develop and evaluate a lung-tissue informed, blood-based multi-omics machine learning model for reliable clinical prediction of emphysema patterns. Timely diagnosis calls for a blood-based predictive model as it may identify emphysema in subjects where CT scans are not clinically indicated. This would also overcome the issues of radiation exposure and false positive findings associated with CT scans. Aim 3 will discover molecularly-informed emphysema subtypes by applying an innovative, interpretable, machine learning algorithm that captures directional feature interactions and provides network representations of the molecular determinants of emphysema subtypes. We will then perform cluster analysis on the Bivariate Shapley network representations to identify distinct subgroups of subjects based on their graph similarity. To confirm the critical regulators of the identified pathways, we will conduct targeted gene silencing and overexpression investigations in airway epithelial cells and lung fibroblasts. Genes will be prioritized for functional validation utilizing existing biological knowledge and network analyses. Through a combination of innovative, cutting-edge data generation, analytic approaches, and functional validation, this project will make a significant contribution by enhancing emphysema phenotyping and multi-omics profiling for a more robust prediction and a better understanding of disease pathobiology. Such knowledge will pave the way for the development of much-needed novel and personalized therapeutic strategies.