Novel artificial intelligence-based approaches to understand the pathological and genetic drivers of primary tauopathies

摘要 Tau通常调节神经元和神经胶质细胞中的微管，但在疾病发病机制中，几个后 翻译修饰会导致这种蛋白质的过度磷酸化，从而对细胞产生毒性。 原发年龄相关的肌萎缩侧索硬化症(部分)，一种与人类衰老相关的常见病理，估计影响 1-7%的人群和患者可以在认知上正常或表现出一系列 症状包括轻度认知障碍或痴呆症。神经病理学上，部分患者有变异性- Ying度神经原纤维缠绕在内侧颞叶，整个没有淀粉样斑块。 大脑。我们的目标是部署三种独立的方法来了解Part如何汇聚和 不同于其他原发和继发性变态反应的特征。目标是使用新的高通量 基因和转录技术与包括计算机在内的创新计算方法相结合 视觉和人工智能，以更好地表征tau磷酸化的部分驱动因素。我们的假设是这台机器 学习分类器(监督和非监督)与单细胞分析相结合将能够准确地 确定和量化转录、基因组、临床和形态特征，部分原因是为了进一步了解 肌萎缩侧索硬化症与淀粉样蛋白无关的潜在机制。我们的理论基础是理解基因 部分转录和临床架构将有助于理解疾病的分期、诊断和 进步。我们计划通过追求以下重要目标来验证我们的假设：(1)对神经原纤维进行量化 使用有监督的机器学习模型处理负担，并将这些数据整合到遗传学和临床病理学中- 临床关联研究(2)模拟神经纤维缠绕变性的序贯进展 无人监督的深度生成性方法。(3)识别与神经原纤维相关的转录改变 部分使用单细胞RNA测序。这项拟议的研究具有创新性，因为它应用了新颖的 转录和机器学习技术在一组研究不足的老年受试者中的识别 缺乏淀粉样变性的肌萎缩侧索硬化。这项拟议的研究具有重要意义，因为它解决了一个关键的未得到满足的需求 开发算法，帮助神经病理学家进行死后诊断，并提供更好的量化分析。 可以帮助促进更好的神经保护策略的标题表型数据。这项建议是建立在 应聘者对与年龄相关的肌萎缩侧索硬化症的明确兴趣，以及他以前接受过的生物医学工程和 翻译基础科学研究。候选人的主要导师约翰·克拉里博士是一位经验丰富的神经科医生- 病理学家和tau神经学家，并将由张斌博士组成的指导团队进行补充， 在计算遗传学和转录组学方面的特殊专业知识和托马斯·福克斯博士，他是 计算病理学领域，具有机器学习分类器、人工智能和计算机方面的专门知识 幻象。他们将保证拟议的研究和培训使申请者成为一名独立的 实验计算神经病理学的研究人员。

ABSTRACT Tau normally regulates microtubules in neurons and glia, however during diseases pathogenesis, several post- translational modifications cause hyperphosphorylation of this protein which consequently is toxic to the cell. Primary age-related tauopathy (PART), a common pathology associated with human aging is estimated to effect 1-7 % of the of the population and patients with the disorder can be cognitively normal or exhibit a range of symptomology including mild cognitive impairment or dementia. Neuropathologically, those with PART have var- ying degrees neurofibrillary tangles in the medial temporal lobe, and an absence of amyloid plaques throughout the brain. Our goal is to deploy three independent approaches to understand how PART has convergent and divergent features from other primary and secondary tauopathies. The objective is to use novel high-throughput genetic and transcriptomic technologies combined with innovative computational methods including computer vision and AI to better characterize drivers of tau phosphorylation in PART. Our hypothesis is that machine learning classifiers (supervised and unsupervised) combined with single cell analysis will be able to accurately identify and quantify transcriptomic, genomic, clinical, and morphological features in PART to further understand the underlying amyloid independent mechanisms of tauopathy. Our rationale is that understanding the genetic transcriptomic and clinical architecture of PART will assistant in understand disease staging, diagnosis, and progression. We plan to test our hypothesis by pursing the following significant aims: (1) Quantify neurofibrillary tangle burden using supervised machine learning models and integrate this data in genetic and clinicopatholog- ical association studies (2) Model the sequential progression of neurofibrillary tangle degeneration in PART with unsupervised deep generative approaches. (3) Identify transcriptional alterations associated with neurofibrillary tangles in PART using single cell RNA sequencing. The proposed research is innovative as it applies novel transcriptomic and machine learning techniques to identify in an understudied group of elderly subjects with tauopathy lacking amyloidosis. This proposed research is significant as it addresses a critical unmet need to develop algorithms which can assist neuropathologists in their post-mortem diagnosis and provide better quan- titative phenotypic data which can aid in facilitating better neuroprotective strategies. The proposal builds upon the candidate's established interest in age-related tauopathy and his prior training in biomedical engineering and translational basic science research. The candidate’s primary mentor, Dr. John Crary, is an experienced neuro- pathologist and tau neuroscientist and will be supplemented by mentoring team consisting of Dr. Bin Zhang with specific expertise in computational genetics and transcriptomics and Dr. Thomas Fuchs, a prominent scientist in the field of computational pathology with a specific expertise in machine learning classifiers, AI, and computer vision. They will assure that the proposed research and training prepare the applicant to be an independent investigator in experimental computational neuropathology.