Implementing a coupled system of integrative ML modeling and data validation for elucidating microglial therapeutic targets in neurodegenerative disease

实施集成机器学习建模和数据验证的耦合系统，以阐明神经退行性疾病中的小胶质细胞治疗靶点

• 批准号: 10699794
• 负责人: Karen SACHS
• 金额: $ 145.73万
• 依托单位: MODULO BIO, INC.
• 依托单位国家: 美国
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-09-15 至 2025-08-31
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10699794
• 关键词: Acceleration; Active Learning; Address; Affect; Antisense Oligonucleotides; Area; Biological; Biological Assay; Biological Models; Biology; Cells; Classification; Clinical; Coculture Techniques; Computer Models; Coupled; Data; Data Set; Databases; Disease; Disease Pathway; Disease Progression; Disease model; Disparate; Drug Targeting; Elements; Etiology; Genes; Goals; Health; Image; In Vitro; Individual; Knowledge; Learning; Link; Methods; Microglia; Modeling; Molecular; Mus; Nerve Degeneration; Neurodegenerative Disorders; Neuroimmune; Neurons; Organism; Pathogenesis; Pathway Analysis; Pathway interactions; Patients; Persons; Physiological; Research; Science; System; Testing; Therapeutic; Therapeutic Intervention; Tissues; Validation; Work; causal model; causal variant; cell type; comparative; data framework; data integration; design; disorder subtype; drug development; effective therapy; flexibility; frontotemporal lobar dementia amyotrophic lateral sclerosis; heterogenous data; in silico; in vitro Model; in vivo; individual patient; induced pluripotent stem cell; insight; large scale data; learning network; machine learning framework; machine learning model; mouse model; multidimensional data; multiple datasets; network models; neuroimmunologic disease; neuroprotection; new therapeutic target; personalized medicine; predictive modeling; risk variant; screening; small molecule; stem cell model; therapeutic target; tool

Project Summary/Abstract: ALS and FTD are fatal neurodegenerative diseases that presently have no cure. To date, one focus area in ALS research has been developing model systems to characterize the condition, with over 20 different ALS mouse models, and more recently, numerous iPSC based models, each gradually contributing to our overall knowledge of the mechanisms behind neurodegeneration, and the contribution of the neuro-immune interface. Despite the multitude of disease models, there is no overarching, computational modeling framework for integrating disparate datasets, towards the goal of characterizing disease networks, and identifying therapeutic targets. Moreover, while standard ML models for target prediction have become ubiquitous in the biomedical sciences, they fail to learn causality, shedding little insight into underlying disease etiology and failing to make effective target predictions. Our proposal’s long-term goal is to create a flexible pipeline, applicable to ND diseases, to characterize the neuro-immune interface and its contribution to ND etiology, to enable therapeutic intervention by creating an integrated workflow to identify ND microglial disease networks in health, disease, and disease subsets. We will capitalize on existing experimental data as well as internal iPSC based in vitro models, paired with a causal ML model. Each component of this workflow can work independently, or can be linked to the other in a powerful ‘active learning’ framework, in which the ML model makes predictions, the co-culture system validates or disproves the prediction, and in each such round the in silico model is refined by integrating the new experimental data. Our causal machine learning model characterizes ND neuro-immune networks from analysis of combined molecular, clinical, and functional data in a multi-layered format with individual layers for ND disease state, data platform, and cell state analyzed simultaneously to bolster confidence for inferences shared among numerous layers and identify unique, and therapeutically relevant, network elements. We will focus initially on therapeutic interventions for ALS, followed by related ND diseases also characterized in the network model. The objectives of this proposal are: (1) to refine an in silico framework for data integration across NDs, microglial subsets, and heterogeneous datasets/data platforms enabling a robust model for therapeutic target prediction and (2) to validate predicted targets in our iPSC microglia and neuron co-culture system using in vitro perturbations (including antisense-oligonucleotides and small molecules) and high-content imaging analysis. The central hypothesis is that comprehensively integrating available data across public datasets and databases, ND diseases, model species, data platforms, and tissue types, with data from our co-culture screening platform, in a powerful mechanistic model, will enable elucidation of causal disease pathways, comparative analysis across conditions, and the identification of therapeutic targets. Ultimately, characterization of individuals can even enable personalized therapy approaches as well as identification of disease subtypes. 1

项目总结/摘要：ALS和FTD是目前无法治愈的致命性神经退行性疾病。 迄今为止，ALS研究的一个重点领域是开发模型系统来表征病情， 有超过20种不同的ALS小鼠模型，最近，许多基于iPSC的模型，每一个都逐渐 有助于我们对神经退行性变背后机制的全面了解， 神经免疫界面尽管有许多疾病模型，但没有一个总体的，计算的 用于整合不同数据集的建模框架，以实现表征疾病网络的目标， 和确定治疗靶点。此外，虽然用于目标预测的标准ML模型已经成为 它们在生物医学科学中无处不在，但它们无法学习因果关系，无法深入了解潜在的疾病 病因学和未能作出有效的目标预测。我们的提案的长期目标是建立一个灵活的 管道，适用于ND疾病，以表征神经免疫界面及其对ND的贡献 病因学，通过创建识别ND小胶质细胞疾病的综合工作流程来实现治疗干预 健康、疾病和疾病子集的网络。我们将利用现有的实验数据， 基于内部iPSC的体外模型，与因果ML模型配对。此工作流的每个组件都可以工作 独立地，或者可以在一个强大的“主动学习”框架中与另一个链接，其中ML模型 进行预测，共同培养系统验证或否定预测，在每一轮这样的循环中， 通过整合新的实验数据，改进了silico模型。我们的因果机器学习模型 通过分析ND患者的分子、临床和功能数据， 一种多层格式，其中包含ND疾病状态、数据平台和分析的细胞状态的各个层 同时加强对在许多层之间共享的推断的信心，并识别唯一的， 治疗相关的网络元素。我们将首先关注ALS的治疗干预， 由相关的ND疾病也在网络模型中表征。本建议的目标是：（1） 完善一个计算机框架，用于跨ND、小胶质细胞亚群和异质性 数据集/数据平台，使得能够建立用于治疗靶点预测的鲁棒模型，以及（2）验证预测的 靶点在我们的iPSC小胶质细胞和神经元共培养系统中使用体外扰动（包括 反义寡核苷酸和小分子）和高含量成像分析。中央 假设是全面整合公共数据集和数据库中的可用数据，ND 疾病、模型物种、数据平台和组织类型，以及来自我们的共培养筛选平台的数据， 一个强大的机制模型，将能够阐明因果疾病的途径，比较分析， 条件和治疗靶点的鉴定。最终，对个体的描述甚至可以 实现个性化治疗方法以及疾病亚型的识别。 1