README: Characterization of the anti-spike IgG immune response to COVID-19 vaccines in people with a wide variety of immunodeficiencies

https://doi.org/10.5061/dryad.6hdr7sr68

This data was collected as part of the PERSIST cohort study (NCT04852276), which aimed to assess the immune response to COVID-19 vaccines in people with immune disorders and healthy volunteers. We collected and analyzed data from April 2021 through April 2022 for this study. Laboratory methods that describe the assay specifics can be found in the associated publication, “Characterization of the anti-spike IgG immune response to COVID-19 vaccines in people with a wide variety of immunodeficiencies” in Science Advances.
 
Five unique datasets were used in our study. Four datasets were derived from participant data:

1. “Final_ELISA_data_v2” provides ELISA measurements of the anti-spike IgG antibody concentration in response to vaccination in all our participants. The file also contains vaccination other clinical information. It is provided in both .csv and .RData formats, though the .RData file is the one that is analyzed by the R code.
2. “Final_MSD_IgG_data” provides electrochemiluminescence measurements of the anti-spike IgG antibody concentration to multiple SARS-CoV-2 variants. It is provided in both .csv and .RData formats, though the .RData file is the one that is analyzed by the R code.
3. “Final_MSD_inhib_data” provides electrochemiluminescence measurements of the ACE2 pseudo-neutralization capacity (% ACE2 inhibition by anti-Spike IgG) to multiple SARS-CoV-2 variants. It is provided in both .csv and .RData formats, though the .RData file is the one that is analyzed by the R code.
4. “Final_AE_data_v3” provides data on post-vaccination adverse events (AE) for all participants in the study. The questionnaire used to collect this data is available at https://doi.org/10.5281/zenodo.8428160. It is provided in both .csv and .RData formats, though the .RData file is the one that is analyzed by the R code.
5. The fifth dataset, “variants.csv”, was created by the CDC and is publicly accessible. We include in our files the copy of the CDC’s dataset that was used for our study. The current version of the CDC’s dataset (which is continuously updated), along with the variable definitions, can be found here: https://data.cdc.gov/Laboratory-Surveillance/SARS-CoV-2-Variant-Proportions/jr58-6ysp.

Description of the data and file structure
 
The variables for all five datasets are defined in the accompanying data dictionary.
 
“Final_ELISA_data_v2” is organized by participant ID (id variable) and the timepoint of their sample (timepoint variable). Variables that remain constant over time such as the participant ID are repeated for each participant across timepoints.
 
“Final_MSD_IgG_data” is organized by participant ID (id variable), the timepoint of their sample (time variable), and the antigen of multiple SARS-CoV-2 variants (antigen variable).
 
“Final_MSD_inhib_data” is organized by participant ID (id variable), the timepoint of their sample (time variable), and the antigen of multiple SARS-CoV-2 variants (antigen variable).
 
“Final_AE_data_v3” is organized by participant ID (id variable) and the relevant COVID-19 vaccine dose (dose variable). Whether the participant experienced any local or systematic adverse event after their vaccine is indicated. What constitutes a local or systemic adverse event is given in the figure legend for Figure S16.
 
The CDC’s “variants.csv” dataset is organized by region (usa_or_hhsregion variable), the date of the Saturday at the end of each week (week_ending variable), and the designation for each SARS-CoV-2 variant of interest (variant variable). Point and interval estimates for each variant’s prevalence in each region and week are provided.
  
Sharing / Access information
 
For Figure 4 of the paper, we used publicly accessible data from the CDC on COVID-19 variants. This data can be found here:
 - https://data.cdc.gov/Laboratory-Surveillance/SARS-CoV-2-Variant-Proportions/jr58-6ysp
 
Code / Software
 
We used R to analyze and visualize all the data in this study. The R script “Code_ELISA_revision1_clean_public_v2” can be used to analyze and visualize the ELISA data related to the anti-spike protein IgG antibody concentrations, adverse events, and breakthrough infections in relation to national SARS-CoV-2 variant proportions. This script thus analyses the following datasets:
·      Final_ELISA_data_v2
·      Final_AE_data_v3
·      variants
 
The R script “Code_MSD_revision1_clean_public_v2” can be used to analyze and visualize the electrochemiluminescence data related to anti-spike protein IgG concentrations and ACE2 inhibition percentage for multiple SARS-CoV-2 variants. Although this script briefly uses the Final_ELISA_data dataset at the end, the script primarily analyzes the following datasets:
·      Final_MSD_IgG_data
·      Final_MSD_inhib_data
 
The R scripts can be found at https://doi.org/10.5281/zenodo.8428160. Note that all errors or warnings elicited by running the R code can be safely ignored. Most occur when R is asked to statistically compare two groups in a multi-panel plot and in at least one panel, such a comparison is not possible. For example, it is not possible to compare the IgG concentrations of healthy volunteers to participants with immunodeficiencies at the 6 months post-dose 3 timepoint because there were no healthy volunteer samples at this time point.
 
Deidentification of the data
 
To limit the possibility of our participants being identified by means of our data, the datasets we include do not report on participants’ age, sex, race, ethnicity, specific immunological conditions, and the collection date of the samples. There were a variety of date variables (e.g., the date of vaccine receipt) that were critical to our analyses. These variables have been retained. However, each date variables’ values were shifted and jittered by different amounts in the datasets we publicly present. Moreover, a different random seed was used before jittering each date variable of interest. While these perturbations do not preserve the statistical quantities of the original data, the statistical issues were outweighed by the need to deidentify data and protect our participants’ privacy. All results we report in the paper, including in the tables and figures, were produced with the original, unperturbed data. The R code for the ELISA data comments on the few specific analyses that are impacted by the shifting and jittering of date variables.