Integrating Machine Learning and Genomic Approaches to Understand Cerebral Small Vessel Disease Pathogenesis from White Matter Hyperintensity Patterns

整合机器学习和基因组方法从白质高信号模式了解脑小血管疾病的发病机制

• 批准号: 10682419
• 负责人: Chia-Ling Phuah
• 金额: $ 17.56万
• 依托单位: WASHINGTON UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-09-20 至 2024-08-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/10682419
• 关键词: 3-Dimensional; Address; Adult; Affect; Alzheimer' s Disease; Amyloid deposition; Archives; Arteries; Arterioloscleroses; Award; Bioinformatics; Biological; Biology; Blood Vessels; Brain; Categories; Cerebral Amyloid Angiopathy; Cerebral hemisphere hemorrhage; Cerebral small vessel disease; Cerebrovascular Disorders; Cerebrovascular Physiology; Clinical; Cluster Analysis; Complex; Computational Technique; Data; Data Set; Dementia; Detection; Development; Development Plans; Disease; Disease Progression; Distal; Elderly; Environment; Ethnic Origin; Etiology; Failure; Functional disorder; Genes; Genetic; Genetic Determinism; Genetic Risk; Genetic Variation; Genomic approach; Genotype; Goals; Grant; Heritability; Heterogeneity; Hypertension; Image; Impaired cognition; Individual; Informatics; Intervention; Investigation; Ischemic Stroke; K-Series Research Career Programs; Knowledge; Lobar; Location; Longitudinal Studies; Machine Learning; Magnetic Resonance Imaging; Maps; Mentorship; Methods; Microvascular Dysfunction; Neurologic; Neurologist; Pathogenesis; Pathogenicity; Pathologic; Pathologic Processes; Pathology; Pathway interactions; Pattern; Phenotype; Play; Population; Prevention; Process; Race; Radiology Specialty; Research; Research Personnel; Research Proposals; Resources; Risk Factors; Role; Statistical Methods; Stroke; Stroke prevention; Testing; Training; Universities; Validation; Variant; Ventricular; Washington; White Matter Hyperintensity; arteriole; biobank; bioinformatics tool; brain parenchyma; career; cohort; design; disorder subtype; effective therapy; genetic analysis; genetic association; genetic variant; genome wide association study; genome-wide; hypertension control; imaging biomarker; imaging genetics; improved; innovation; insight; interdisciplinary approach; machine learning method; medical schools; multidisciplinary; nervous system disorder; neuroimaging; new therapeutic target; novel; patient population; population based; programs; risk variant; segregation; serial imaging; stroke therapy; success; therapeutic development; tool; treatment strategy; unsupervised learning

PROJECT SUMMARY As a neurointensivist and neurologist at Washington University School of Medicine in St. Louis (WUSM), my career goal is to develop an independent research program as a computational biologist capable of using advanced bioinformatics and statistical methods to integrate analysis of large-scale neuroimaging and genetic data, with the aim of deepening understanding of the biological mechanisms influencing cerebral small vessel disease (SVD) and identifying new targets for therapeutic development. As a first step towards this goal, I have designed an innovative proposal that combine machine-learning (ML) methods and integrated imaging genetic analyses of large-scale neuroimaging and genetic data to improve characterization of SVD disease mechanisms. The clinical, imaging, and etiologic heterogeneity of SVD have impeded efforts to uncover the pathophysiology of this common and debilitating neurological disease. White matter hyperintensities (WMH), a major imaging endpoint of SVD, are comprised of multiple SVD pathologic processes. Growing evidence suggests location-specific vulnerability of brain parenchyma to different underlying SVD pathologic processes, in which spatially localized WMH patterns may reflect distinct SVD etiologies. Characterizing WMH spatial pattern variations in SVD will not only provide insights into underlying pathogenesis, such as vascular amyloid deposition, arteriolosclerosis, and other less well defined or as-yet unknown disease mechanisms, but also lead to creation of novel imaging biomarkers of these SVD pathologic processes. This proposal addresses a key inadequacy, as existing WMH pattern definitions are determined empirically and cannot distinguish overlapping SVD etiologies and risk factors. In this proposal, I aim to capture WMH spatial pattern variations that reflect distinct SVD etiologies in an unbiased manner, by applying clustering analysis/ML methods to structural MRI data to create novel etiology-specific SVD imaging phenotypes. Moreover, given that genetics influence the variance of WMH, I will integrate genetic analyses of these WMH patterns to uncover novel mechanisms that influence SVD pathogenesis. My preliminary data demonstrate the feasibility of identifying data-driven WMH spatial pattern variations, which are specific to distinct SVD etiologies, and allow detection of genetic risk variants that may help inform SVD pathologic processes. My career plan leverages the extensive resources and exceptional environments at WUSM, under the guidance of a multidisciplinary mentorship team with expertise across diverse fields including cerebrovascular physiology, neuroimaging, informatics, genetics, and machine learning (Drs. Jin-Moo Lee, Daniel Marcus, Carlos Cruchaga and Yasheng Chen). In this Career Development Award, I propose to: 1) determine distinct WMH spatial patterns that can discriminate underlying SVD pathology and/or risk factors by applying pattern analysis ML methods to structural MRI data from three unique cohorts (n=2,710) enriched for different SVD pathologies (Aim 1a), and examining if the ML-defined WMH patterns segregate individuals by well-defined SVD risk factors as biologic validation (Aim 1b), and 2) identify genetic variants (Aim 2a) associated with WMH patterns that reflect diverse pathologic processes influencing SVD using genome wide association and gene-based analyses; replicate the top variants (Aim 2b) in an independent population-based cohort (n=21,708); and use advanced bioinformatics tools to uncover new biologic pathways associated with WMH spatial patterns (Aim 2c). This research proposal and accompanying development plan with focused training in machine learning, neuroimaging, and multivariate methods for integrated imaging genetics analysis, will build on my background in genetics towards a career investigating cerebrovascular disorders using translational bioinformatics. This Award will provide me with the necessary training to evolve into an independent investigator with a computational research program that can integrate large imaging and genetics datasets to derive results that are highly relevant to the prevention and treatment of cerebrovascular disease in my clinical patient population.

项目总结 作为圣路易斯华盛顿大学医学院(WUSM)的一名神经强化专家和神经学家，我的 职业目标是作为一名计算生物学家开发一个独立的研究计划，能够使用 先进的生物信息学和统计方法集成大规模神经成像和基因分析 数据，目的是加深对影响脑小血管的生物学机制的了解 疾病(SVD)和确定治疗发展的新靶点。作为迈向这个目标的第一步，我已经 设计了一种将机器学习(ML)方法和集成成像遗传技术相结合的创新方案 对大规模神经成像和遗传数据的分析，以改善SVD疾病机制的特征。 SVD的临床、影像和病因学的异质性阻碍了揭示 这种常见的、使人衰弱的神经疾病的病理生理学。白质高强度(WMH)，a SVD的主要影像终点由多个SVD病理过程组成。越来越多的证据 提示脑实质对不同潜在的SVD病理过程具有特定位置的脆弱性， 其中空间局域化的WMH模式可以反映不同的SVD病因。描述WMH空间 SVD的模式变化不仅将提供对潜在发病机制的洞察，如血管淀粉样蛋白 沉积、动脉硬化和其他不太清楚或尚不清楚的疾病机制，但也导致 创建这些SVD病理过程的新的成像生物标记物。这项建议解决了一个关键问题 不充分，因为现有的WMH模式定义是根据经验确定的，无法区分重叠 Svd的病因和危险因素。在这项建议中，我的目标是捕捉反映WMH空间模式变化的 通过将聚类分析/ML方法应用于结构MRI，以无偏倚的方式识别SVD病因 创建新的病因学特定的SVD成像表型的数据。此外，考虑到遗传因素对 WMH的变异，我将整合这些WMH模式的遗传分析，以揭示新的机制 影响SVD发病机制。我的初步数据证明了识别数据驱动的WMH的可行性 空间模式变异，专用于不同的SVD病因，并允许检测遗传风险变异 这可能有助于了解SVD的病理过程。 我的职业生涯规划利用了WUSM的广泛资源和特殊环境，在 拥有包括脑血管在内的不同领域的专业知识的多学科指导团队的指导 生理学、神经成像、信息学、遗传学和机器学习(李振武博士、丹尼尔·马库斯博士、卡洛斯博士 Cruchaga和Yaseng Chen)。在这个职业发展奖中，我建议：1)确定不同的WMH 通过应用模式分析可以区分潜在的SVD病理和/或风险因素的空间模式 ML方法对三个不同SVD病理丰富的独特队列(n=2,710)的MRI数据进行结构化 (目标1a)，并检查ML定义的WMH模式是否通过明确定义的SVD风险因素将个体隔离 作为生物验证(目标1b)，以及2)识别与WMH模式相关联的遗传变异(目标2a 使用全基因组关联和基于基因的分析来反映影响SVD的不同病理过程； 在一个独立的基于人群的队列(n=21,708)中复制顶级变种(Aim 2b)；并使用高级 生物信息学工具，以发现与WMH空间模式相关的新生物途径(目标2c)。 这项研究建议和伴随着机器学习中的重点训练的发展计划， 神经成像和用于综合成像遗传学分析的多变量方法将建立在我的背景之上 利用翻译生物信息学研究脑血管疾病的遗传学方向。这 获奖将为我提供必要的培训，使我成为一名具有计算能力的独立调查员 可以集成大型成像和遗传学数据集以得出高度相关的结果的研究计划 在我的临床患者群体中预防和治疗脑血管疾病。