Model-based inference and forecasting of co-circulating pathogen dynamics

项目摘要 公共卫生持续面临来自多种病原体的威胁，但病原体 与不同疾病相关的疾病通常在监测方面被划分， 管理和研究。这种划分的方法忽略了许多方面， 病原体相互作用，在某些情况下导致它们对人类的集体负担加重。 公共卫生这些相互作用可以是生物学的（例如，交叉反应性免疫），行为 (e.g.,促进对良好卫生的坚持），或临床（例如，误诊）。现代，数据- 驱动的数学建模方法有可能解决合作的动态， 通过解释这些相互作用来传播病原体。在这样做的时候，建模也有 通过借用信息，有可能改善病原体特异性疾病预测 不同疾病的监测数据。迄今为止，这一潜力在很大程度上仍未得到开发。在这 项目，我将开发一个通用的框架，用于建模的动态共循环 病原体该框架的第一个组成部分将使用贝叶斯分层建模， 将病原体传播动力学的机械描述与 监控过程，允许最大限度地利用异构数据流， 为生物学推论提供信息。这个框架的第二个组成部分将涉及验证 通过对未来疾病动态的预测进行模型推断。这两个组成部分 框架将涉及使用多个模型，这些模型代表了关于 病原体相互作用，以及其他形式的模型不确定性。该框架将 应用于两种情况：巴西的蚊媒病毒和印第安纳州的呼吸道病原体。在 在这两种情况下，最近出现的病原体和地方性病原体的共同传播提出了新的 监视和控制活动的挑战，使新的建模工具的发展 应对这些挑战尤为及时。

PROJECT SUMMARY Public health faces threats from a multitude of pathogens on an ongoing basis, yet pathogens associated with different diseases are typically compartmentalized with respect to surveillance, management, and research. This compartmentalized approach ignores the many ways that pathogens interact, in some cases leading to the exacerbation of their collective burden on public health. These interactions can be biological (e.g., cross-reactive immunity), behavioral (e.g., prompting adherence to good hygiene), or clinical (e.g., misdiagnosis). Modern, data- driven approaches to mathematical modeling have the potential to resolve the dynamics of co- circulating pathogens by accounting for these interactions. In doing so, modeling also has the potential to improve pathogen-specific disease forecasts by borrowing information across surveillance data for different diseases. To date, this potential remains largely untapped. In this project, I will develop a generalizable framework for modeling the dynamics of co-circulating pathogens. The first component of this framework will use Bayesian hierarchical modeling to fuse mechanistic descriptions of pathogen transmission dynamics with statistical descriptions of surveillance processes, allowing for maximal leveraging of heterogeneous data streams to inform biological inferences. The second component of this framework will involve validating model inferences through forecasts of future disease dynamics. Both components of this framework will involve the use of multiple models that represent competing hypotheses about pathogen interaction, as well as other forms of model uncertainty. This framework will be applied in two settings: mosquito-borne viruses in Brazil and respiratory pathogens in Indiana. In both of these settings, co-circulation of recently emerged and endemic pathogens poses new challenges for surveillance and control activities, making the development of new modeling tools to address these challenges especially timely.