Identifying Temporal Sequences of Exposures in Type 1 Diabetes Etiology

项目总结/摘要 1型糖尿病（T1 DM）是一种慢性自身免疫性疾病，其中不能产生 胰岛素导致葡萄糖调节异常。T1 DM患者经历了显著的疾病负担， 预防疾病是优先事项。环境暴露与免疫激活和进行性免疫有关。 破坏产生胰岛素的β细胞，导致明显的T1 DM。尽管有几项研究调查了 在T1 DM病因学中的环境暴露中，尚未确定致病性暴露。一个原因是 对时间和组合的理解有限，即，时间序列的暴露， T1 DM病因。修改这些风险可能提供安全和具有成本效益的方法，以防止或延迟 T1 DM疾病发作。因此，更清楚地了解曝光的时间顺序， 需要T1 DM病因学为预防策略提供信息并减少疾病负担。 目前的预防策略侧重于改变T1 DM自然史中的早期疾病时间点， 包括免疫激活和进行性β细胞破坏。免疫激活使用胰岛 自身抗体和β细胞破坏使用症状前T1 DM分期来近似。时间 对这些早期疾病时间点的暴露顺序知之甚少。因此本 一项提案旨在确定通过调节胰岛β细胞功能来增加T1 DM风险的暴露时间顺序， 自身抗体轨迹和症状前分期的进展。本提案的具体目标 是1）。确定共同改变胰岛自身抗体轨迹的暴露时间序列是否增加 T1 DM的风险，以及2.）确定与T1 DM阶段相关的暴露时间序列 可以预测T1 DM的症状前进展。这项工作将使用最先进的时间 机器学习和数据挖掘方法与糖尿病的环境决定因素的数据， Young（TEDDY）研究。本项目的成功完成将提高曝光的临床可译性 通过进一步了解暴露的时间和组合，改进T1 DM预防 与早期T1 DM疾病时间点有关。本提案中开发的分析管道将 也可以作为一个框架，阐明复杂的时间相互作用，在其他疾病驱动的疾病。 本申请概述了犹他州大学严格的科学和临床培训计划。 时间机器学习、联合轨迹建模和序列模式挖掘中的计算训练 生物医学信息学系将与该部门的纵向临床经验相结合 儿科学和内分泌学这些活动，增加了来自科学和临床的指导 专家，将使申请人成为一个成功的医生，科学家在儿科内分泌学。

PROJECT SUMMARY/ABSTRACT Type 1 diabetes mellitus (T1DM) is a chronic autoimmune disease in which an inability to produce insulin results in glucose dysregulation. Patients with T1DM experience significant disease burden, making disease prevention a priority. Environmental exposures are implicated in immune activation and progressive destruction of insulin-producing β-cells that lead to overt T1DM. Despite several studies investigating the role of environmental exposures in T1DM etiology, no causative exposures have been identified. One reason is limited understanding of the timing and combination, i.e., temporal sequences, of exposures that contribute to T1DM etiology. Modifying these exposures may offer safe and cost-effective approaches to prevent or delay T1DM disease onset. Therefore, a clearer understanding of the temporal sequences of exposures that drive T1DM etiology is needed to inform prevention strategies and reduce disease burden. Current prevention strategies focus on altering early disease timepoints in the natural history of T1DM, including immune activation and progressive β-cell destruction. Immune activation is monitored using islet autoantibodies, and β-cell destruction is approximated using presymptomatic T1DM staging. The temporal sequences of exposures underlying these early disease timepoints are poorly understood. Therefore, this proposal seeks to identify temporal sequences of exposures that increase risk for T1DM by modulating islet autoantibody trajectories and progression through presymptomatic staging. The specific aims of this proposal are to 1.) determine if temporal sequences of exposures that jointly alter islet autoantibody trajectories increase risk for T1DM, and 2.) determine if temporal sequences of exposures that are associated with stages of T1DM can predict presymptomatic progression of T1DM. This work will be investigated using state-of-the-art temporal machine learning and data mining methods with data from the Environmental Determinants of Diabetes in the Young (TEDDY) study. Successful completion of this project will improve the clinical translatability of exposure modification in T1DM prevention by providing further insight into the timings and combinations of exposures that are implicated in early T1DM disease timepoints. The analytical pipelines developed in this proposal will also serve as a framework for elucidating complex temporal interactions in other exposure-driven diseases. This application outlines a rigorous scientific and clinical training plan at the University of Utah. Computational training in temporal machine learning, joint trajectory modeling, and sequential pattern mining in the Department of Biomedical Informatics will be combined with longitudinal clinical experiences in the Division of Pediatrics and Endocrinology. These activities, augmented with mentorship from scientific and clinical experts, will enable the applicant to become a successful physician-scientist in pediatric endocrinology.