Advanced therapeutic hypothermia efficacy network modeling in neonatal HIE

新生儿 HIE 的先进低温治疗功效网络模型

基本信息

批准号：
10538972
负责人：
ALLEN D EVERETT
金额：
$ 71.58万
依托单位：
JOHNS HOPKINS UNIVERSITY
依托单位国家：
美国
项目类别：
财政年份：
2022
资助国家：
美国
起止时间：
2022-09-02 至 2026-07-31
项目状态：
未结题

来源：
https://reporter.nih.gov/project-details/10538972
关键词：
2 year old Address Adult Age Alabama Automobile Driving Biological Biological Markers Brain Injuries Caring Cessation of life Child Clinical Clinical Data Communities Computer Simulation Coupled Data Data Scientist Data Set Data Sources Effectiveness Enrollment Funding Future Health system Individual Intervention Intervention Trial Investigation Knowledge Life Machine Learning Magnetic Resonance Imaging Manuscripts Measures Methods Modeling Molecular Molecular Profiling National Institute of Child Health and Human Development Neonatal Neonatal Intensive Care Units Nucleic Acids Outcome Pathway Analysis Patient Selection Patients Peer Review Probability Proteins Publishing Research Personnel Retinoscopy Right to Treatments Risk Factors Severities Stratification Testing Therapeutic Time Universities Validation Virginia base brain magnetic resonance imaging clinical heterogeneity clinical phenotype cohort cost data-driven model disability exosome experience feeding follow-up improved improved outcome in silico individual patient machine learning method modifiable risk molecular marker natural hypothermia neonatal encephalopathy neonatal hypoxic-ischemic brain injury neonate network models neural network neurodevelopment novel novel marker patient safety phenotypic biomarker phenotypic data predicting response predictive modeling prevent prospective protein metabolite response safety outcomes socioeconomics

项目摘要

Fifty percent of neonatal encephalopathy cases result from hypoxic-ischemic encephalopathy (HIE). Therapeutic hypothermia (TH), the only approved therapy remains state of the art care for HIE, despite only a 30% reduction in death and significant disability. Our limited ability to accurately track TH efficacy limits identification of babies, who may benefit from adjunctive therapies. Under R01HD086058, our team enrolled neonates with HIE treated with TH and tested whether circulating brain injury biomarkers used in adults were associated with HIE severity, MRI and 2-year outcomes. We identified the novel biomarkers significantly associated with the proposed outcomes and published 22 peer-reviewed original, high-impact manuscripts. Our team has extensive experience in biomarkers in children (1R01HL150070), brain injury biomarkers in HIE (U01 NS114144) and real-time machine learning integrating within health systems (R61HD105591). Our central hypothesis is that a holistic and integrative approach, including deep clinical and community-based data, and molecular biomarkers of multiple biologic pathways, analyzed using a fully connected parsimonious neural network will best describe relationships with longitudinal outcomes, and be able to predict response to TH in individual patients. Our outstanding group of investigators from Johns Hopkins University, University of Virginia and University of Alabama Birmingham, propose the following Aims: Aim 1a. Perform clinical data- driven modeling to ascertain TH effectiveness. We will use deep phenotyping data sets of all maternal, neonatal, community-based, and follow-up data collected retrospectively (2016-2021) and prospectively thru year 1, from neonates treated with TH at the 3 centers (n = 500) to model TH efficacy using multivariable methods against longitudinal outcomes. Aim 1b. Identify novel molecular signatures for HIE insult severity which predict response to TH. Using our discovery (N=178) TH treated HIE cohort, we will determine if circulating brain injury proteins, metabolites and exosome proteins and nucleic acids are associated with TH efficacy. Aim 1c. Determine relationships emerging from integration between clinical, community-based, and molecular markers using a fully connected parsimonious neural network approach. 1C.1 Use computational simulations to identify the levers, modifiable risk factors and interventions associated with the probability of negative outcomes, in the neural network, and 1C.2 Determine in silico whether optimization of the neural network using those levers at the individual patient level, results in a reduction in the predicted probability of negative outcomes. Aim 2. External validation of neural network and estimation of potential clinical gain achievable by optimization of the neural network, in prospective patients (Years 2-5). Completion of our aims will identify the clinical, socioeconomic, and molecular mechanisms driving clinical heterogeneity in HIE and response to TH. We will then be poised to rapidly deploy a dynamic, precision-based model to optimized patient selection for future HIE adjunctive therapies.

百分之五十的新生儿脑病病例是由缺氧缺血性脑病 (HIE) 引起的。治疗性低温 (TH) 是唯一被批准的治疗方法，尽管目前只有死亡和严重残疾减少 30%。我们准确跟踪 TH 功效极限的能力有限识别可能受益于辅助治疗的婴儿。在R01HD086058下，我们的团队注册了患有 HIE 的新生儿接受 TH 治疗，并测试了成人中使用的循环脑损伤生物标志物是否有效与 HIE 严重程度、MRI 和 2 年结果相关。我们显着地鉴定了新的生物标志物与拟议成果相关，并发表了 22 篇经过同行评审的高影响力原创手稿。我们的团队在儿童生物标志物（1R01HL150070）、HIE脑损伤生物标志物方面拥有丰富的经验（U01 NS114144）和卫生系统内集成的实时机器学习（R61HD105591）。我们的中心假设是一种整体和综合的方法，包括深入的临床和基于社区的方法使用完全连接的简约模型对数据和多种生物途径的分子生物标志物进行分析神经网络将最好地描述与纵向结果的关系，并能够预测对个别患者的 TH。我们来自约翰·霍普金斯大学、美国大学的杰出研究人员团队弗吉尼亚州和阿拉巴马大学伯明翰分校提出以下目标：目标 1a。执行临床数据- 驱动建模以确定 TH 有效性。我们将使用所有母体的深度表型数据集，回顾性（2016-2021 年）和前瞻性收集的新生儿、社区和随访数据第 1 年，从在 3 个中心接受 TH 治疗的新生儿 (n = 500) 中使用多变量建立 TH 疗效模型针对纵向结果的方法。目标 1b。识别 HIE 损伤严重程度的新分子特征预测对 TH 的反应。利用我们的发现 (N=178) TH 治疗 HIE 队列，我们将确定是否循环脑损伤蛋白、代谢物以及外泌体蛋白和核酸与 TH 相关功效。目标 1c。确定临床、社区和社区之间的整合所产生的关系使用完全连接的简约神经网络方法进行分子标记。 1C.1 使用计算模拟以确定与发生概率相关的杠杆、可修改的风险因素和干预措施神经网络中的负面结果，以及 1C.2 在计算机上确定是否优化神经网络在个体患者层面使用这些杠杆的网络，会导致预测概率的降低负面结果。目标 2. 神经网络的外部验证和潜在临床收益的估计在未来的患者（2-5 年）中，可以通过优化神经网络来实现。完成我们的目标将确定导致 HIE 和临床异质性的临床、社会经济和分子机制对TH的反应。然后，我们将准备快速部署动态的、基于精度的模型来优化未来 HIE 辅助治疗的患者选择。