Norovirus correlates of protection

This is the public repository for the paper titled, "Serum antibodies are unreliable correlates of protection against norovirus infection and disease," by Miller et al.

This study examines the relationships between antibody titers and clinical outcomes in two human challenge studies (data are located here and here).

Organization

• The general setup of this repository is as follows:

  • The assets folder contains the reference files accessed in document rendering (i.e. main text and supplement)

  • The code folder contains all code used to run analyses and generate results.

  • The data folder contains the final processed data used in model fitting.

  • The products folder contains all files for submission.

  • The results folder contains all model output, results, figures, and tables generated by any of the code scripts.

Reproducibility Instructions

• These analyses are set up as an R project. To reproduce this analysis, you will need the R and RStudio software, as well as the packages cited and listed in the Session Info in the Supplement. The project uses renv to help ensure reproducibility and contains all needed packages.

• To start, open the .Rproj file with RStudio.

• Run renv::restore() to download all used packages and dependencies in the versions used in conducting this analysis. This will install the cmdstanr package along with all others. After this installation, follow the steps listed on this post to complete installation and prepare for model fitting: Getting started with CmdStanR

• The project folder and sub-folders contain README files that provide further information. Each R script is also fully documented to allow easy reproducibility.

• To completely replicate our main analysis and recreate figures and tables shown in the main text and supplement, run the code scripts in the following order:

1)  `code/01-mod-fit-main-analysis.R`

2)  `code/02-fig-gen-main-analysis.R`

3)  `code/03-tbl-gen-main-analysis.R`

• Code scripts starting with 04-06 and 07-09 run the associated code for sensitivity analyses using different limit of detection definitions and not accounting for censoring, respectively. Script 10-tbl-gen-combined-results.R uses results aggregated in scripts 03, 06, and 09 to create one overall table organized by model version.

• The model-functions.R (or model-functions-no-censoring.R for that tsensitivity analysys) script will be sourced at the beginning of the model fitting scripts, so it does not need to run explicitly. However, it contains all the code used to fit the models and extract all needed results for plotting and table generation, so any alterations would need to take place in that script. Stan models will be compiled at the beginning of every script where needed, and can be found in the code/stan-code folder.

• As many of the object and file names are the same across different versions of the analysis (but separated by subfolders), I recommend clearing your environment in between running different versions of the analysis.

• General LoD figure and table generation is located in code/11-lod-figures.R.

License

All of the code we wrote is distributed under the Affero GNU GPL v3.0. Any code that was not written by us has been attributed and is released under the conditions of its own license. Any text and images we created are licensed under the Creative Commons 4.0 BY-NC-SA international license.