Hardening Software for Rule-based models-Competitive Revision

点击翻译按钮获取中文摘要

PROJECT SUMMARY/ABSTRACT In this competitive revision application, we are proposing to expand the scope of Research Project 2R01GM111510-05 by adding a new sub-aim to Specific Aim 3. As originally formulated, the goal of Aim 3 was to apply new features of PyBioNetFit (PyBNF) in modeling studies of immunoreceptor signaling. This activity now becomes Aim 3a. The new sub-aim, Aim 3b, will be focused on data-driven modeling of the effects of vac- cination and immunity-evading SARS-CoV-2. The modeling of Aim 3b will complement Aims 1 and 2 by driving improvements of PyBNF that will be broadly useful for epidemiological modelers. Aim 3b addresses a need for situational awareness, i.e., an ability to monitor for signs of new surges in incidence of severe COVID-19. Aim 3b also addresses a need to monitor for waning of natural and vaccine-induced immunity and emergence of new strains of SARS-CoV-2 that are capable of evading vaccine-induced immunity. This work will extend our recently published COVID-19 forecasting efforts in which we used mathematical models for region-specific COVID-19 epidemics to make accurate short-term predictions of COVID-19 case detection. In this work, we focused on making predictions for metropolitan areas, which are defined on the basis of socioeconomic coher- ence. We have found that metropolitan areas are more uniformly impacted by COVID-19 than states. Most forecasting to date has focused on making state-level predictions vs. predictions for cities and their sur- rounding metropolitan areas. We plan to extend our existing models to account for vaccination in the 15 most populous metropolitan statistical areas (MSAs) in the United States. After new versions of these region- specific models are formulated, we will begin to update model parameterizations daily using Bayesian infer- ence. Daily updates are important for maintaining prediction accuracy and for modifying the models to account for changes in social-distancing behaviors. Our daily inferences will include quantification of forecast uncertain- ties, so as to allow for detection of surges and confident rapid responses. The model structure that we are us- ing as the basis for our forecasts is a deterministic compartmental model that extends the classic SEIR model, which consists of four ordinary differential equations (ODEs) for the dynamics of susceptible (S), exposed (E), infected (I), and removed (R) populations. Our extended model accounts for a) the variable time from infection to onset of symptoms, which is non-exponentially distributed; b) shedding of virus by asymptomatic individuals; c) mild and severe forms of symptomatic disease; d) quarantine driven by testing and contact tracing; and e) widespread implementation of time-varying social-distancing measures. Here, we are proposing to extend the model further to account for vaccination, including vaccines that require booster shots and the time required for development of vaccine-induced immunity. We will also develop models in which persons with immunity be- come susceptible gradually over time to currently circulating variants of SARS-CoV-2 and models that account for emergence of immunity-evading variants.