Personalized Rejection Risk Assessment in Cardiac Transplantation

项目概述：心脏移植排斥反应(CAR)是移植医学中一个严重关注的问题，代表着 短期和长期同种异体移植物存活的主要威胁。因此，汽车监控和预防是一项 移植后护理的主要重点，受者经常接受定期监测的心肌内膜检查- Dial活检(EMB)用于组织学分级，以及频繁、计划的免疫抑制降级- 锡安(IS)。这种标准化的汽车减排方法的一致性是由于无法使用 基于个人汽车风险的可靠、主动和量身定制的策略。因此，汽车风险较低的患者 暴露于不必要的EMB程序和过度的IS治疗，而高危患者经历 不充分的车辆监控和IS的早期/不适当断奶。这使患者面临潜在的伤害， 并强调了对精确的汽车风险评估工具的明确、未得到满足的需求。这件事的首要前提是 建议是包含在临床数据和EMB组织中，作为常规护理的一部分，已在 移植中心有办法提供可行的汽车风险评估。广泛的免疫学，诊断 诊断和药理学数据被捕获在移植中心的电子健康记录(EHR)中，虽然 EMB组织学样本的集合存储在病理档案中(通常是数字化的)。这项提议旨在 利用先进的机器学习算法和现场诊断方法来深入挖掘这些档案 资源，用于验证新的汽车风险预测模型。在目标1中，我们将利用我们的经验- 使用自动组织分析流水线开发用于预测未来医疗的“形态模型” 使用已存档的H&E幻灯片。苏木精伊红(H&E)组织学幻灯片是从所有EMB事件中生成的 作为标准护理的一部分。在已发表和获得专利的先前工作中，我们提取了定量形态 数字化H&E幻灯片中的特征，在建模时，显示出卓越的诊断性能 心肌损伤和汽车等级。在目标2中，我们将超越标准的H&E，利用我们在 移植的eMBS的定量原位免疫分析以开发用于预测的“形态分子”模型 未来的汽车。这一目标将在激动人心的试点工作的基础上扩展，这些试点工作展示了汽车风险分层的潜力 将定量图像分析与多重免疫荧光相结合。最后，在目标3中，我们将开发一个 将AIMS 1和AIMS 2的数据与综合临床数据相结合预测CAR的组织信息学模型 从EHR中提取的信息学数据。最终，作为这项工作的结果，我们预计将验证一种新的预 用于探索个性化汽车监控和预调查的前瞻性调查的精度预测模型 发明策略。除了潜在的翻译影响，这项研究计划将建立在申请人的 具有复杂队列设计、综合数据建模和移植免疫诊断学的知识。与.一起 计划的课程工作和多样化的指导、咨询和协作团队，这项建议提供了选项- MAL是佩斯特博士成长为一名在多模式诊断方面拥有成熟专业知识的调查员的工具。

Project Summary: Cardiac allograft rejection (CAR) is a serious concern in transplant medicine, representing the leading threat to short- and long-term allograft survival. As a result, CAR surveillance and prevention is a primary focus of post-transplant care, with recipients undergoing frequent, scheduled, surveillance endomyocar- dial biopsy (EMB) for histologic CAR grading along with frequent, scheduled de-escalation of immunosuppres- sion (IS). The uniformity of this standardized approach to CAR mitigation is the result of an inability to employ reliable, proactive, and tailored strategies based on individual CAR risk. Consequently, patients at low CAR risk are exposed to unnecessary EMB procedures and excess IS therapy, while patients at high risk experience inadequate CAR surveillance and early/inappropriate weaning of IS. This exposes patients to potential harm, and highlights the clear, unmet need for precision CAR risk-assessment tools. The overarching premise for this proposal is that contained within the clinical data and EMB tissues already collected as part of usual care at transplant centers exists the means to provide actionable CAR risk assessments. Extensive immunologic, diag- nostic, and pharmacologic data are captured in electronic health records (EHR) at transplant centers, while large collections of EMB histology samples are stored (and often digitized) in pathology archives. This proposal seeks to utilize advanced machine-learning algorithms and in-situ diagnostic methods to deeply mine these archival resources for the purpose of validating novel CAR risk-prediction models. In Aim 1, we will leverage our experi- ence with automated histologic analysis pipelines to develop a ‘morphologic model’ for predicting future CAR using archived H&E slides. Hematoxylin-and-Eosin (H&E) histology slides are generated from all EMB events as part of standard-of-care. In published and patented prior efforts, we have extracted quantitative morphologic features from digitized H&E slides which, when modeled, demonstrate excellent performance for diagnosing myocardial injury and CAR grades. In Aim 2, we will move beyond standard H&E, leveraging our experience with quantitative, in-situ immune-profiling of transplant EMBs to develop a ‘morpho-molecular’ model for predicting future CAR. This aim will expand upon exciting pilot work which showed the CAR risk-stratification potential of combining quantitative image-analysis with multiplex immunofluorescence. Finally, in Aim 3, we will develop a ‘histo-informatics’ model for predicting CAR by integrating data from Aims 1 & 2 with comprehensive clinical informatics data extracted from the EHR. Ultimately, as a result of this work, we expect to validate a novel pre- cision prediction model for use in prospective investigations exploring personalized CAR surveillance and pre- vention strategies. Beyond the potential translational impact, this research plan will build on the Applicant’s knowledge of complex cohort design, integrated data modeling, and transplant immunodiagnostics. Along with planned coursework and a diverse mentoring, advisory, and collaborative team, this proposal provides the opti- mal vehicle for Dr. Peyster’s maturation into an investigator with proven expertise in multi-modality diagnostics.