Drug Discovery for Alzheimer’s Disease Enabled by Multi-Omics and Artificial Intelligence

通过多组学和人工智能实现阿尔茨海默病药物发现

• 批准号: 10301220
• 负责人: Jesse Meyer
• 金额: $ 23.4万
• 依托单位: MEDICAL COLLEGE OF WISCONSIN
• 依托单位国家: 美国
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-09-01 至 2023-04-30
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/10301220
• 关键词: Aging; Alzheimer' s Disease; Alzheimer' s disease brain; Alzheimer' s disease pathology; Alzheimer' s disease therapeutic; American; Amyloid beta-Protein; Area; Artificial Intelligence; Biological Assay; Biological Models; Biology; Brain; Brain region; Cells; Chemical Structure; Chemicals; Data; Disease Progression; Drug Screening; Glutamates; Goals; Health; Human; In Vitro; Induced pluripotent stem cell derived neurons; Knowledge; Learning; Libraries; Literature; Maps; Mission; Molecular; N-Methylaspartate; Nerve Degeneration; Neurodegenerative Disorders; Neuronal Dysfunction; Neurons; Outcome; Partner in relationship; Pathway interactions; Pharmaceutical Preparations; Pharmacotherapy; Phenotype; Physiological; Prevention approach; Process; Proteome; Proteomics; Public Health; Research; Research Personnel; Resistance; Resources; Stress; Synapses; System; Testing; Therapeutic; Therapeutic Intervention; Toxic effect; Training; United States National Institutes of Health; Work; abeta accumulation; abeta oligomer; abeta toxicity; base; biological research; cost; drug candidate; drug discovery; drug structure; effective therapy; human data; human disease; human stem cells; improved; in silico; in vitro Model; in vivo; induced pluripotent stem cell; innovation; metabolomics; multiple omics; neuron loss; neuropathology; neurotoxicity; novel therapeutics; palliate; palliative; prevent; preventive intervention; programs; relating to nervous system; resilience; therapy development

PROJECT SUMMARY/ABSTRACT There is a fundamental gap in our understanding of how amyloid beta oligomers (AβO) induce neurotoxicity and neuron death in Alzheimer’s disease (AD), as evidenced by a dearth of therapies to prevent or halt AD progression. Continued existence of this knowledge gap represents a major issue for public health and the mission of the NIH because, until it is filled, development of treatments for neurodegeneration in AD will remain largely intractable. The long-term goal of this work is to discover pathways that enable resistance to AβO- induced neurotoxicity thereby allowing discovery of new AD therapeutics. The overall objective here, which is the next step in pursuit of this goal, is to build AI that accurately predicts the ability of drug candidates to cure or prevent toxicity of AβO in human stem cell-derived cortical glutamatergic neurons. To train this AI, a library of proteomic and metabolomic (hereafter referred to as multi-omic) phenotypes will be generated from neurons that are: 1) healthy, 2) AβO-treated (AD-like), or 3) drug library+AβO-treated. The central hypothesis is that some drugs at least partially palliate AβO-induced neurotoxicity, which is observable as a shift in multi-omic state toward the healthy state, and that AI can learn to predict this curative potential from drug structures. This hypothesis is based on preliminary data generated by the applicant and literature. The rationale for the proposed research is that mapping the difference in multi-omic phenotypes of healthy and AβO-stressed neurons, and mapping how chemical structures induce changes between those states, will allow AI to learn to make accurate predictions of whether additional, unmeasured molecules can improve neuron health. This will result in new and innovative approaches for prevention and treatment of AD. Guided by preliminary data and literature, this hypothesis will be tested by pursuing two specific aims: 1) validate the multi-omic phenotype landscape of healthy and AD-like neurons; and 2) build AI to discover new drugs that prevent AβO-induced neuron death in AD. The first aim will validate the human disease relevance of our model system using cell- based assays and by comparing omic profiles from our system to those observed in human AD brains. The second aim will build a map of how drugs candidates alter neural multi-omic states to use for training predictive AI. Completion of these aims will contribute (1) an in vitro system that mimics physiological milieu, and also (2) molecular ‘omics’ signatures of those healthy and AD-like human iPSC-derived neural cells, which are two areas of high program relevance defined in NOT-AG-19-007. This approach is innovative, in the applicant’s opinion, because it departs from the status quo by using highly translatable human iPSC-derived neurons for unbiased discovery of palliative drug candidates using a unique combination of multi-omics and AI. This contribution will be significant because it is expected to vertically advance our understanding of basic neuron stress resistance, as well as result in the first drugs that prevent AβO neural toxicity. Ultimately, such knowledge will be useful for other neurodegenerative disorders of aging.

项目摘要/摘要 在我们对淀粉样β寡聚体(AβO)如何引起神经毒性的理解上存在着根本的差距 以及阿尔茨海默病(AD)中的神经元死亡，缺乏预防或阻止AD的治疗方法证明了这一点 进步。这种知识差距的持续存在是公共卫生的一个重大问题， NIH的使命，因为在它被填满之前，AD神经退行性疾病的治疗方法的开发将继续 在很大程度上很难处理。这项工作的长期目标是发现能够抵抗βO-的途径。 诱导神经毒性，从而使发现新的AD疗法成为可能。这里的总体目标是 追求这一目标的下一步，是建立能够准确预测候选药物治愈能力的人工智能 或阻止AβO对人类干细胞来源的皮质谷氨酸能神经元的毒性。为了训练这个人工智能，一个图书馆 蛋白质组和代谢组(以下称为多组体)表型将从神经元产生 即：1)健康，2)AβO处理(类AD)，或3)药库+AβO处理。中心假设是 一些药物至少部分缓解了A-βO诱导的神经毒性，这是一个多组学上的转变 状态朝向健康状态，人工智能可以学习从药物结构预测这种治疗潜力。这 假设是基于申请者产生的初步数据和文献。该计划的基本原理 建议的研究是定位健康的和AβO-应激的多组表型的差异 神经元，以及绘制化学结构如何在这些状态之间引起变化的地图，将使人工智能能够学习 准确预测额外的、未测量的分子是否可以改善神经元健康。这将是 导致预防和治疗阿尔茨海默病的新的和创新的方法。根据初步数据和 文献中，这一假说将通过追求两个具体目标来验证：1)验证多组体表型 健康和AD样神经元的景观；以及2)建立人工智能以发现预防AβO诱导的新药 阿尔茨海默病的神经元死亡。第一个目标是验证我们的模型系统与人类疾病的相关性，使用细胞- 基于分析，并将我们系统的基因组图谱与在人类AD大脑中观察到的基因组图谱进行比较。这个 第二个目标是建立一张地图，展示候选药物如何改变神经多组体状态，以用于训练预测性 艾。这些目标的完成将有助于(1)模拟生理环境的体外系统，以及(2) 这些健康的和类似AD的人类IPSC来源的神经细胞的分子组学特征，这是两个 NOT-AG-19-007中定义的与计划高度相关的领域。这种方法是创新的，在申请人的 观点，因为它偏离了现状，使用高度可翻译的人类IPSC来源的神经元来 使用多组学和人工智能的独特组合，不偏不倚地发现缓解药物候选。这 贡献将是巨大的，因为它有望垂直地推进我们对基础神经元的理解 抗应激能力，以及第一批预防AβO神经毒性的药物。最终，这样的 这些知识将对其他衰老的神经退行性疾病有用。