Public repository for the Boni modeling group's paper: "SARS-CoV-2 epidemic after social and economic reopening in three US states reveals shifts in age structure and clinical characteristics."
Authors: Nathan B Wikle, Thu Nguyen-Anh Tran, Ephraim M Hanks, Maciej F Boni, and 12 others.
This repository contains source code and additional supplementary materials from our manuscript, "SARS-CoV-2 epidemic after social and economic reopening in three US states reveals shifts in age structure and clinical characteristics." This manuscript has been accepted for publication in Science Advances; a preprint is available on medRxiv.
The following instructions provide details on how to run the source code underlying the analysis, including replication of the main figures and results.
The code has been tested with R version 3.6.0 (2019-04-26) -- "Planting of a Tree." The following R packages (version number in parentheses) must be installed before the code will run successfully:
mc2d(0.1.18)mvtnorm(1.1.1)readxl(1.3.1)fda(5.1.9)snow(0.4.3)doParallel(1.0.16)foreach(1.5.1)msm(1.6.8)splines2(0.4.1)reticulate(1.18)doMC(1.3.7)extrafont(0.17)scales(1.1.1)dplyr(1.0.7)forcats(0.5.1)viridis(0.6.1)hrbrthemes(0.8.0)ggplot2(3.3.5)ggridges(0.5.2)
In addition, the Ubunto Mono font family is used in the production of several figures. It can be downloaded here.
This repository contains two main directories, ./Aug2021 and ./Nov2020. These folders contain supplementary materials for the accepted version of this manuscript and an earlier preprint, respectively.
The ./Aug2021 directory contains several subfolders:
cpp-v5-discharges-nonhospdeaths: C++ source code to perform a deterministic solve of our age-structured ordinary differential equations (ODE) model.data: CSVs containing observed epidemic data from Rhode Island, Massachusetts, and Pennsylvania.inference: R source code to reproduce the analysis and figures in the paper/supp. materials.output: output files from our analysis and figures found in the manuscript and supp. materials.
In addition, this repository contains a main file (./Aug2021/inference/main.R) which will (a) replicate the analysis in the paper and supplementary materials, and (b) create Figures 2 and 5 from the manuscript as well as supplementary figures S3-S17. The script contains 4 sequential steps, which perform the following:
- Before running any code, make sure the required R packages have been installed. Set the R working directory to the location of the
main.Rfile (ie,./Aug2021/inference).
- Load required packages into R.
- Compile
ODESIM(deterministic solution to ODE model) from C++ source code.
-
Replicate the analyses from the manuscript and supplementary materials. Note that this step is extremely computationally intensive, and is commented out by default. The sourced R scripts recreate the inference procedures used to obtain all results in the manuscript and supplementary materials, including the "best fitting models" shown in Figures 2, 5, and S13 and S14, as well as the model comparison discussed in Table 3, the sensitivity analyses in Supp. Materials (SM) Section 5.1 (time from symptoms to hospitalization), SM Section 5.2 (time from symptoms to presentation), and SM Section 6 (excess deaths). The output from these scripts are stored in a directory, "./output/MCMC-redo", that is automatically created when the file is souced. Note that output from these files is already included in "./output/manuscript-results"; it is not necessary that these files be sourced if interest is only in the existing results.
-
If you choose to recreate all inferential results, it is HIGHLY RECOMMENDED that these files be run in PARALLEL. A single R script, distributed across five 2.2 GHz Intel Xeon processors from the PSU high-performance computing infrastructure (PSU Roar Supercomputer), takes ~2.5 days to run. If run sequentially, this section of the analysis would take ~65 days.
- Create Figures 2 and 5 and S3-S17 from source code. The results are saved as PDFs or PNGs in
./Aug2021/output/figures.