emilianbold / SAPHIRE

Modelling transmission dynamics of COVID-19, while accounting for presymptomatic infectiousness, time-varying ascertainment rates, transmission rates and population movements.

Modelling transmission dynamics of COVID-19

This repository includes data and codes for reproducing the results in the manuscript:

Hao X, Cheng S, Wu D, Wu T, Lin X, and Wang C (2020). Reconstruction of the full transmission of COVID-19 in Wuhan. Nature, doi: 10.1038/s41586-020-2554-8.

In the following sections, we will describe the purposes of each major directory and the scripts in that folder. One should keep in mind that the scripts and data here are meant for reproduction of the results in the paper only. Scripts were tested on R-3.6.x.

We encourage you to raise purely technical questions through Github issues so that we can answer your questions ASAP.

Prerequisite third-party R packages

You may need to install the following R packages if you have not done so yet:

• BayesianTools
• coda
• cairoDevice
• vioplot
• readr
• corrplot
• IDPmisc

Descriptions of folders, scripts and other files

Scripts in folders scripts_main, scripts_resurgence and scripts_sensitivity are meant to be run directly, while scripts in other folders are supportive and not meant to be run directly by users.

Folder scripts_main

R scripts for our main analyses. Since we constantly need to use functions defined in R directory and previous outputs located at output directory, it is necessary to set the code_root variable at the beginning of the script properly. code_root should be set to the directory under which this README file is located. code_root should ends with /. The same applies to other scripts that are meant to be directly run by users. For example, suppose you have git cloned SAPHIRE at /home/Sarah/SAPHIRE and this Readme.md file is at /home/Sarah/SAPHIRE, then code_root variable should be set to /home/Sarah/SAPHIRE/.

• Run_SEIR_main_analysis.R: R script to reproduce the main analyses (Fig. 2). This R script will call SEIRfitting function to perform the analysis. After the run, please first inspect output/par_traj_run_main_analysis.png visually to make sure the MCMC run has converged. If convergence has not been achieved, rerun the script with different random seeds, or specify a good initial parameter values.

• confirm_convergence.R: Confirm and test convergence of MCMC by comparing three independent runs. This script will reproduce Supplementary Fig. 10. This requires output from Run_SEIR_main_analysis.R . Run Run_SEIR_main_analysis.R first before running this script.

Folder scripts_resurgence

R scripts for risk of resurgence estimations.

• Run_resurge_simulation.R: R script to run the risk of resurgence estimations after control measures is lifted using the parameters from the main and sensitivity s8 analyses. This requires output from scripts_main/Run_SEIR_main_analysis.R and scripts_sensitivity/Run_SEIR_s8.R. Run Run_SEIR_main_analysis.R and scripts_sensitivity/Run_SEIR_s8.R first before running this script.

• Run_resurge_plot_fig3_A.R: R script to reproduce Fig. 3A. This requires output from scripts_main/Run_SEIR_main_analysis.R and Run_resurge_simulation.R.

• Run_resurge_plot_fig3_BC.R: R script to reproduce Fig. 3B/C. This requires output from scripts_sensitivity/Run_SEIR_s8.R and Run_resurge_simulation.R.

Folder R

This folder contains major R functions used for our model fitting, parameter estimation and producing results figures.

• fun_SEIRpred.R: Function SEIRpred evolves the system according to deterministic model specified by Eqs. 1-7.

• fun_SEIRsimu.R: Function SEIRsimu evolves the system according to the stochastic model specified by Eqs. 10-16.

• fun_SEIRfitting.R : SEIR model fitting and parameter estimation by MCMC with the Delayed Rejection Adaptive Metropolis (DRAM) algorithm.

• fun_SEIRresurge.R: Resurgence simulation using the SIER model

• fun_SEIRplot.R: The plot function for the figures of main and sensitivity analyses

• fun_R0estimate.R : The function to calculate the $R_0$ (basic/effective reproduction number)

• fun_Findzero.R: Function Findzero, given previous parameter estimation results and corresponding initial conditions, will find dates of zero case, i.e., date when I=0 and when E+P+I+A=0.

• init_cond.R: Function generate_init_condi creates a list containing all parameters, useful constants, initial conditions of the population set according to the parameters, and several functions that need to be accessed in various R functions.

• correlationPlot_modified.R: A modified version of correlationPlot in package BayesianTools.

Folder scripts_sensitivity

R scripts for out sensitivity analyses (s1-s9).

• Run_SEIR_s1.R: R script to run the sensitivity analyses s1: Adjust the reported incidences from January 29 to February 1 to their average.
• Run_SEIR_s2.R: R script to run the sensitivity analyses s2: Assume an incubation period of 4.1 days (lower 95% CI from reference 2) and presymptomatic infectious period of 1.1 days (lower 95% CI from reference 10 is 0.8 days, but our discrete stochastic model requires $D_p>1$), equivalent to setting $D_e=3$ and $D_p=1.1$, and adjust P(0) and E(0) accordingly.
• Run_SEIR_s3.R: R script to run the sensitivity analyses s3: Assume an incubation period of 7 days (upper 95% CI from reference 2) and presymptomatic infectious period of 3 days (upper 95% CI from reference 10), equivalent to set $D_e=4$ and $D_p=3$, and adjust P(0) and E(0) accordingly.
• Run_SEIR_s4.R: R script to run the sensitivity analyses s4: Assume the transmissibility of the presymptomatic and unascertained cases is $α=0.46$ (lower 95% CI from reference 15) of the ascertained cases.
• Run_SEIR_s5.R: R script to run the sensitivity analyses s5: Assume the transmissibility of the presymptomatic and unascertained cases is $α=0.62$ (upper 95% CI from reference 15) of the ascertained cases.
• Run_SEIR_s6.R: R script to run the sensitivity analyses s6: Assume the initial ascertainment rate is $r_0=0.14$ (lower 95% CI of the estimate using Singapore data) and adjust A(0), P(0), and E(0) accordingly.
• Run_SEIR_s7.R: R script to run the sensitivity analyses s7: Assume the initial ascertainment rate is $r_0=0.42$ (upper 95% CI of the estimate using Singapore data) and adjust A(0), P(0), and E(0) accordingly.
• Run_SEIR_s8.R: R script to run the sensitivity analyses s8: Assume the initial ascertainment rate is $r_0=1$ (theoretical upper limit) and adjust A(0), P(0), and E(0) accordingly.
• Run_SEIR_s9.R: R script to run the sensitivity analyses s9: Assume no unascertained cases by fixing $r_0=r_{12}=r_3=r_4=r_5=1$.

Folder data

This folder contains the main data used in the study. Detailed description of the data can be found in the following paper:

Pan A, Liu L, Wang C, Guo H, Hao X, Wang Q, Huang J, He N, Yu H, Lin X, Wei S, Wu T (2020). Association of Public Health Interventions With the Epidemiology of the COVID-19 Outbreak in Wuhan, China. JAMA, 323(19):1915-1923.

• Covid19CasesWH.csv: This file contains daily counts of laboratory-confirmed cases with onset between Dec 8, 2019 and Mar 8, 2020.

Folder output

This folder stores the results of parameters estimations and the fitting plots.