ATD: Collaborative Research: Multi-task, Multi-Scale Point Processes for Modeling Infectious Disease Threats

该项目开发新的基于点过程的算法，用于建模和预测事件级传染病数据，例如当流行病出现、接近消除时，或用于接触者追踪。通过该项目制定的方法可用于查明超级传播者事件的源头、确定病例输入趋势以及更好地评估新出现的流行病和未来的大流行病。通过该项目开发的方法还可以应用于流行病学之外的点过程，包括社交媒体、地震学和犯罪学。该项目将培养两名统计学和计算机科学方面的博士生。该项目将在三年的资助期内，每年资助一所大学的研究生。该项目开发了新的基于点过程的算法，用于解决在一系列时间和空间尺度上建模传染病威胁时出现的四个重要任务：1)结合现实的传播和报告机制；2)在连接受监测的独立地理区域的传播图中进行链接预测；3)对超级传播者事件的时空和网络位置进行源检测；4)在几十年时间尺度和全球空间尺度上对新出现的疾病流行进行建模。通过期望最大化算法推导出一种可用于接触跟踪和源检测的概率分支结构。制定多元霍克斯过程来推断跨不同地理区域的交叉传播，其中需要新的理论和方法来处理超过临界阈值1的繁殖。通过该项目开发了类似于隔间模型的点过程，可以纳入实际的传播和漏报机制（例如接触期、无症状病例），以改进预测和流行率估计。最后，本项目开发新出现的流行病事件的模型，以确定疾病参数相对于空间和时间的可分离性，并评估爆发成为大流行的风险。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This project develops new point-process based algorithms for modeling and forecasting event-level infectious disease data, such as when an epidemic is emerging, near elimination, or for contact tracing. The methods developed through the project have applications to source detection of super-spreader events, identification of case importation trends, and providing better risk assessments of emerging epidemics and future pandemics. The methods developed through the project also have applications beyond epidemiology where point processes are used, including social media, seismology, and criminology. The project will train two PhD students in statistics and computer science. This project will support one graduate student per year at each university for each of the three years of the grant. This project develops new point-process based algorithms for solving four important tasks that arise in modeling infectious disease threats over a range of temporal and spatial scales: 1) incorporating realistic transmission and reporting mechanisms, 2) link prediction in the transmission graph connecting separate geographic regions under surveillance, 3) source detection of the spatial-temporal and network locations of super-spreader events, and 4) modeling emerging disease epidemics over timescales of decades and spatial scales of the globe. Expectation maximization algorithms are derived to infer a probabilistic branching structure that can be used for contact tracing and source detection. Multivariate Hawkes processes are formulated to infer cross-transmission across separate geographic regions, where new theory and methods are needed to handle reproduction above the critical threshold of 1. Point process analogs to compartmental models are developed through the project that can incorporate realistic transmission and under-reporting mechanisms (e.g. exposure period, asymptomatic cases) to improve forecasts and prevalence estimation. Finally, this project develops models of emerging epidemic events for determining the separability of disease parameters vs. space and time and assessing the risk that an outbreak will become a pandemic.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.