Reclassifying Pulmonary Arterial Hypertension Into Immune Phenotypes Using Machine Learning

这是安德鲁·斯威特博士的K23奖项申请，他是一名肺部/重症监护内科医生和年轻的研究员 斯坦福大学，他正在建立一个在肺动脉高压(PAH)精确度方面的利基市场 表型鉴定。他的工作重点是使用机器学习对PAH进行重新分类，检测隐藏的模式 在高通量分子数据中发现新的表型。现有的PAH临床分类没有 告知治疗决定，结果总体上是糟糕的，采用“一刀切”的治疗方法。有一个 迫切需要分子表型工作，以开发更接近病理生物学的分类方案 并确定可用于治疗的患者亚群。斯威特博士的K23建立在创新的基础上 他使用机器学习根据血液免疫图谱对PAH患者进行分组的研究，没有 根据临床特点进行指导。这种不可知的方法发现了4种具有不同细胞因子的免疫表型 独立于临床亚型并对疾病风险分层的配置文件。这些发现表明， 炎症是PAH重新分类的可行平台。广泛的研究表明，炎症与 PAH和多重免疫靶向疗法正在积极研究中，但这些研究依赖于 假设存在共同的病理表型。斯威特博士的K23的目标是更好地理解 多环芳烃免疫表型的纵向进化、机制基础和治疗 这意味着什么。首先，他将在两个观察队列(美国斯坦福大学、谢菲尔德大学、 英国)在病程期间重新评估免疫表型(目标1)。基于初步数据，动态 一些患者可能会发生表型转换，并反映出临床疾病严重程度的变化。接下来，他会 集成血液转录分析和应用复杂的计算工具，以提供特定表型 机械论的见解(目标2)。他推测，不同的转录图谱将把表型与特定的 信号通路和免疫细胞亚群。研究结果将使用来自公众的多队列数据进行验证 储存库。最后，他将在最近两个免疫的PAH试验队列中进行后组细胞因子分析 对调节剂进行了测试，以评估不同表型的治疗反应是否有所不同(目标3)。他的研究可能 帮助识别对特定治疗有反应的患者，告知临床试验设计，导致生物标记物 发现并定义了多环芳烃中的新生物学。K23将为斯威特博士提供所需的关键支持 过渡到独立的研究生涯，并成为多环芳烃精确表型的领导者。他的K23目标 将在PAH临床表型/队列建设方面获得经验，扩大生物信息学专业知识，培养 合作，并将发现转化为R01发展的新假设。他将由一个忠诚的人来指导 多学科导师团队(罗汉姆·扎马尼亚[多环芳烃临床试验设计/生物标记物专家]，Marlene Rabinovitch[翻译多环芳烃研究的领导者]和Purvesh Khatri[生物信息学的先驱])和 顾问(Mark Nicolls[翻译多环芳烃免疫学]、PJ Utz[免疫学]和Manisha Desai[生物统计学])。

This is a K23 award application for Dr. Andrew Sweatt, a pulmonary/critical care physician and young investigator at Stanford University who is establishing a niche in pulmonary arterial hypertension (PAH) precision phenotyping. His work centers on using machine learning to reclassify PAH, where hidden patterns are detected in high-throughput molecular data to uncover new phenotypes. The existing PAH clinical classification does not inform therapy decisions, and outcomes are overall poor with a ‘one-size-fits-all’ treatment approach. There is a critical need for molecular phenotyping efforts, to develop classification schemes that sit closer to pathobiology and identify therapeutically-targetable patient subsets. Dr. Sweatt’s K23 builds on an innovative foundational study where he used machine learning to cluster PAH patients based on blood immune profiling, without guidance from clinical features. This agnostic approach uncovered 4 immune phenotypes with distinct cytokine profiles that are independent of clinical subtypes and stratify disease risk. These findings indicate that inflammation is a viable platform for PAH reclassification. Extensive research has implicated inflammation in PAH and multiple immune-targeting therapies are under active investigation, but these studies rest on the assumption that a common pathophenotype exists. The objective of Dr. Sweatt’s K23 is to better understand PAH immune phenotypes in terms of their longitudinal evolution, mechanistic underpinnings, and therapeutic implications. First, he will perform serial cytokine profiling in two observational cohorts (Stanford, USA; Sheffield, UK) to reassess immune phenotypes during the disease course (Aim 1). Based on preliminary data, dynamic phenotype switches may occur in some patients and reflect changes in clinical disease severity. Next, he will integrate blood transcriptomic profiling and apply sophisticated computational tools to provide phenotype-specific mechanistic insights (Aim 2). He postulates that distinct transcriptomic profiles will link phenotypes to specific signaling pathways and immune cell subsets. Findings will be validated using multi-cohort data from public repositories. Finally, he will perform post-hoc cytokine profiling in two recent PAH trial cohorts where immune modulators were tested, to assess if therapy responses differ across phenotypes (Aim 3). His research could help identify patients who will respond to specific therapies, inform clinical trial designs, lead to biomarker discovery, and define novel biology in PAH. The K23 will provide Dr. Sweatt with the critical support needed to transition to an independent research career and be a leader in PAH precision phenotyping. His K23 objectives are to gain experience in PAH clinical phenotyping/cohort building, expand expertise in bioinformatics, cultivate collaboration, and translate findings to new hypotheses for R01 development. He will be guided by a committed team of multidisciplinary mentors (Roham Zamanian [expert in PAH clinical trial design/biomarkers], Marlene Rabinovitch [leader in translational PAH research], and Purvesh Khatri [pioneer in bioinformatics]) and scientific advisors (Mark Nicolls [translational PAH immunology], PJ Utz [immunology], and Manisha Desai [biostatistics]).