RAPID: COVID-19 Transmission Network Reconstruction from Time-Series Data

点击翻译按钮获取中文摘要

Covid-19 has spread rapidly since it was detected in Hubei province in China. To estimate potential impact of Covid-19, researchers have employed models to predict numbers of people infected, and the potential morbidity caused by the virus. Importantly, results from models guide policies for controlling the spread of the COVID-19 virus, where it is also recognized that conclusions reached vary considerably based on the model being employed. Here, for effective mitigation and prediction of the spread of the virus it is important to construct the transmission network of COVID-19 which informs the routes the virus takes in introducing or re-introducing infections to different regions and populations. An accurate estimate of the transmission network will help in developing models with higher fidelity and accuracy and can help in effective mitigation strategies. The project will develop a data-driven approach for reconstruction of transmission network of COVID-19, to complement and aid model-based approaches. Here, relative interdependence and independence of infection in a region from other infections in other regions estimated solely from data history will be employed. Such a data-driven approach has the potential to evolve significant complementary insights and guide strategies for COVID-19 mitigation.There are numerous parametric models being employed to analyze/predict evolution of the Covid-19 based viral infection. In this project, the focus is to unravel the evolution of the transmission network of infections as inferred from data. A primary approach is based on filtering and multivariate optimal estimation. The first step of the methodology is to identify the pathway by which an agent can affect another agent where all intermediate agents that facilitate transmission of infection from one agent to another are identified. In the second step the focus is on estimating the dynamics of the transmission whereby aspects such as the delay in expression of infection from the time the agent encounters an infected agent are unraveled. Tools from graphical models and their relationship to filtering over networks will be brought to bear on the problem. The data-driven algorithms are agnostic to models bringing complimentary set of insights into the Covid-19 transmission from the model-based approaches currently being employed. The filtering/optimization-based methods will be used on data generated by standard epidemiological models such as the Susceptible-Infected-Removed models.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.