Repository for the paper "Within-host viral growth and immune response rates predict FMDV transmission dynamics for African Buffalo"

Abstract

Infectious disease dynamics operate across biological scales: pathogens replicate within hosts but transmit among populations. Functional changes in the pathogen-host interaction thus generate cascading effects across organizational scales. We investigated within-host dynamics and among-host transmission of three strains (SAT1, 2, 3) of foot-and-mouth disease viruses (FMDVs) in their wildlife host, African buffalo. We combined data on viral dynamics and host immune responses with mathematical models to ask (i) How do viral and immune dynamics vary among strains?; (ii) Which viral and immune parameters determine viral fitness within hosts?; and (iii) How do within-host dynamics relate to virus transmission? Our data reveal contrasting within-host dynamics among viral strains, with SAT2 eliciting more rapid and effective immune responses than SAT1 and SAT3. Within-host viral fitness was overwhelmingly determined by variation among hosts in immune response activation rates but not by variation among individual hosts in viral growth rate. Our analyses investigating across-scale linkages indicate that viral replication rate in the host correlates with transmission rates among buffalo and that adaptive immune activation rate determines the infectious period. These parameters define the virus's relative basic reproductive number ($\mathcal{R}_0$), suggesting that viral invasion potential may be predictable from within-host dynamics.

Authors

Joshua C. Macdonald, Hayriye Gulbudak, Brianna Beechler, Erin E. Gorsich, Simon Gubbins, Eva Perez-Martin, and Anna E. Jolles

Corresponding authors contact

• AJ: aejolles-AT-gmail.com, HG: h.gulbudak-AT-louisiana.edu

Contact for questions about program files

• JCM: joshuamac-AT-tauex.tau.ac.il (temporary, postdoc email), or macdonald.j.caleb-AT-gmail.com (stable)

Software version, package, and license information

All code in this repository is available under a Creative Commons International 4.0 license with attribution. Authors wishing to modify this code for their own purposes should cite the version of this work archived in Zenodo. The MATLAB scripts in this repository use only base MATLAB install modules and were written using release R2020b. We have tested the code with release R2023b and found no compatibility issues. The Juypter Notebook was originally written with Python 3.6.11. We have tested it with Python version 3.10.12 as packaged by conda-forge and found no compatibility issues. The packages used in this notebook are numpy, scipy, pandas seaborn, scikit-learn, and matplotlib. When testing with python 3.10.12 we used versions 1.26.0, 1.11.3, 2.1.3, 0.13.0, 1.3.2, 3.8.2 of these packages respectively and found no compatibility issues.

The files

Folders in this repository are:

1. Finalized fitting
2. Output analysis

Finalized fitting

The Finalized fitting folder contains the (i) finalized code for model fitting as described in the methods section and SI of the paper, (ii) the input experimental data formatted to work with this code, (iii) the code to generate parameter confidence intervals for both the needle and contact-infected hosts, (iv) the code to generate the time plots shown in figures 2 and S3 of the manuscript, (v) output parameters in 3D (host x paramter x simulation number) .mat format. needed to run the CI generation and plot generation files (vi) For useers that may not have access to matlab there are 2D csv versions of these files included. We note this code is written in an object oriented style, and so most code will require no modifications from the end user. The only file which an end user may wish to modify is Get_Params.m which is a demonstration script showing how one can use the backend program files to generate a specified number of Monte Carlo simulations / parameter estimates for each individual host. Specific files in this folder (with brief description) are:

Data files

Input data

• HaptoDataContact.csv contains the Haptoglobin time series data (a measure of innate immune response) for each of the twelve contact-infected hosts. Rows correspond to each host, and columns to observation day (in terms of time since contact was initiated). Units are micrograms per mL (log base 10 is taken prior to fitting).
• VNTContactData.csv contains the Viral neutralization (VNT) time series data for each of the twelve contact-infected hosts. Rows correspond to each host, and columns to observation day (in terms of time since contact was initiated). Units are log base 10 VNT.
• ViremiaContact.csv contains the Viral load time series data afor each of the twelve contact-infected hosts. Rows correspond to each host, and columns to observation day (in terms of time since contact was initiated). Units are log base 10 genome copies per mL.
• HaptoDataNeedle.csv contains the Haptoglobin time series data (a measure of innate immune response) for each of the twelve needle-infected hosts. Rows correspond to each host, and columns to observation day (in terms of time since experiment start). Units are micrograms per mL (log base 10 is taken prior to fitting).
• VNTNeedleData.csv contains the Viral neutralization (VNT) time series data for each of the twelve needle-infected hosts. Rows correspond to each host, and columns to observation day (in terms of time since experiment start). Units are log base 10 VNT.
• ViremiaNeedle.csv contains the Viral load time series data afor each of the twelve needle-infected hosts. Rows correspond to each host, and columns to observation day (in terms of time since experiment start). Units are log base 10 genome copies per mL.

Output data

• InfectionStartTImes.csv contains the posterior infection start time estimates generated in our previous work (https://doi.org/10.5281/zenodo.5121203)
• params_noise.mat is the 3D matlab matrix file containing the 10,000 paramter estimates per host used to generate figure 2 of the manuscript
• params_noise.csv is the 2D csv version of the matlab matrix file of the same name to allow access to indivdual host parameter estimates for those without matlab (Warning - large file, may crash Excel)
• params_noise_needle.mat is the 3D matlab matrix file containing the 10,000 paramter estimates per host used to generate figure S3 of the manuscript
• params_noise_needle.csv is the 2D csv version of the matlab matrix file of the same name to allow access to indivdual host parameter estimates for those without matlab (Warning - large file, may crash Excel)

For comparision across scales

• QualCompInd.csv has median paramter estimates for both within and between-host (see https://doi.org/10.5281/zenodo.5121203) parameters across all contact infected hosts. Fever quantities are from https://doi.org/10.1186/s13567-022-01076-3

Program files

Support files that don't require end user modification

• ContactDataHaptoStimes.m is the matlab code to generate a single instance of the Monte-Carlo simulations for a specified contact-infected host
• NeedleDataHaptoStimes.m is the matlab code to generate a single instance of the Monte-Carlo simulations for a specified needle-infected host
• GetData.m is the file to generate a new time series given a model fit
• ViralGrowth.m generates initial estimates of the initial viral load and viral growth rate as described in the manuscript
• GetDataInitial.m generates new inital data points for immune response data given a drawn infection start time as described in the manuscript

Scripts with function calls to replicate figures, and tables given the generated parameter sets in the data files

• tplotsNew.m generates the time plots presented in figure 2
• tplotsNewNeedle.m generates the time plots presented in figure S3
• CIsNew.m generates the CIs and means displayed in the SI for contact infected hosts
• CIsNewNeedle.m generates the CIs and means displayed in the SI for needle infected hosts

Demonstration script go generate a new set of parameter estimates per host

• GetParams.m generates the paramter estimates over 10,000 simulations for all hosts (both needle and contact infected) by calling support function files

Output analysis

Output analysis contains the mean parameter estimates for each serotype over the 10,000 Monte-Carlo simulations per host as well as the python script that generates the correlation and regression analyses as well as figures 3-5 of the manuscript.

Data files

• FMDVMeansMay9.csv 10,000 sample means per serotype, columns are individual parameters, rows are individual fitting interations
• QualCompInd.csv has median paramter estimates for both within and between-host parameters across all contact infected hosts

Program files

• Fitting_analysis.ipynb Python notebook for correlation and regression analyses as well as generation of figures 3-5.