Development of a Precision Medicine-based Diagnostic Tool for Membranous Nephropathy

摘要/摘要 这个项目的目标是开发一种精确的医学方法来快速诊断膜性疾病。 肾病(MN)使用从肾脏活检获得的蛋白质组数据的自动统计分析。这 该方法使用数据独立获取质谱仪(DIA-MS)和算法数据流水线 能够有效地确定肾活检组织中存在的最可能的MN抗原类型。锰是一种 大多数情况下由循环致病原引起的异质性自身免疫性肾病 自身抗体与足细胞抗原反应，导致致病性的形成和积聚 肾小球毛细血管环周围的免疫复合体。以PLA2R型MN为例，确定了 抗原类型已被证明对诊断、监测治疗反应和早期发现很重要 疾病爆发的证据。在历史上，MN抗原类型的确定一直是通过免疫染色进行的； 然而，由于至少发现了17种抗原类型，这已经变得不切实际。往往是不够的 活检样本中的组织进行这一数量的免疫染色，而且免疫染色过程 既耗费时间又耗费资源。DIA-MS的使用为抗原分型提供了一种新的蛋白质组学方法 免疫复合体通过从冰冻活检组织中洗脱捕获，消化成胰蛋白酶多肽， 然后用DIA-MS进行测量。使用算法分类识别候选MN抗原，然后 在最后的免疫染色步骤中验证以确认候选抗原。我们的初步研究表明 这是一种健壮的方法；然而，如果没有类似的健壮的数据分析管道，该方法就不能扩展。 在这个第一阶段的项目中，我们将优化DIA-MS方法，然后从已知案例中收集定量数据 可用于开发、训练、测试和优化算法分类的最常见的MN类型 使用机器学习(ML)方法的模型。为了训练ML模型，我们将收集DIA-MS蛋白质 来自PLA2R、THSD7A和Exostosin类型的MN各50个样本以及50个样本的丰富数据 对这些抗原均为阴性的人作为对照。在第二阶段，我们将为所有用户构建完整的数据集 MN的已知抗原类型，并针对诊断工作流优化ML分类器模型。成功完成 这些目标将导致开发一种有效地对任何MN病例进行分类的综合方法 抗原类型。这些工具将在很大程度上推动肾脏病理学的实践。 从疾病诊断到基于精确医学的蛋白质组学方法，将有效地提供可行的 为护理MN患者的临床医生提供信息。

Summary/Abstract The goal of this project is to develop a precision medicine approach to the rapid diagnosis of membranous nephropathy (MN) using automated statistical analysis of proteomic data obtained from kidney biopsies. This approach uses data-independent acquisition mass spectrometry (DIA-MS) and an algorithmic data pipeline capable of efficiently determining the most likely MN antigen types present in kidney biopsy tissue. MN is a heterogenous autoimmune kidney disease that is caused in most cases by the presence of circulating pathogenic autoantibodies that react with podocyte antigens leading to the formation and accumulation of pathogenic immune complexes around glomerular capillary loops. Using the example of PLA2R-type MN, determination of antigen type has been shown to be important for diagnosis, monitoring response to treatment and early detection of disease flares. Historically, determination of MN antigen type has been performed by immunostaining; however, this has become impractical due to the discovery of at least 17 antigen types. There often is not enough tissue in the biopsy sample to conduct this number of immunostains, and moreover the immunostaining process is both time and resource intensive. The use of DIA-MS provides a novel proteomics approach to antigen typing in which immune complexes are captured by elution from frozen biopsy tissue, digested into tryptic peptides, and then measured by DIA-MS. Candidate MN antigens are identified using algorithmic classification and then validated in a final immunostaining step to confirm the candidate antigen. Our preliminary studies indicate that this is a robust approach; however, the method is not scalable without a similarly robust data analysis pipeline. In this Phase I project, we will optimize the DIA-MS method and then collect quantitative data from known cases of the most common types of MN that can be used to develop, train, test and optimize algorithmic classification models using a machine learning (ML) approach. In order to train the ML models, we will collect DIA-MS protein abundance data from 50 samples each of PLA2R, THSD7A and Exostosin types of MN, as well as 50 samples that are negative for each of these antigens as controls. In the Phase II, we will build complete datasets for all known antigen types of MN and optimize the ML classifier model for diagnostic workflows. Successful completion of these aims will result in the development a comprehensive method to efficiently classify MN cases of any antigen type. These tools will advance the practice of renal pathology from a largely morphology-based approach of diagnosing disease to a precision medicine-based proteomics approach that will efficiently provide actionable information to clinicians caring for patients with MN.