OpenABM-Covid19: Agent-based model for modelling the Covid-19 epidemic

Description

OpenABM-Covid19 is an agent-based model (ABM) developed to simulate the spread of Covid-19 in a city and to analyse the effect of both passive and active intervention strategies. Interactions between individuals are modelled on networks representing households, work-places and random contacts. The infection is transmitted between these contacts and the progression of the disease in individuals is modelled. Instantaneous contract-tracing and quarantining of contacts is modelled allowing the evaluation of the design and configuration of digital contract-tracing mobile phone apps.

A full description of the model can be found here, a dictionary of input parameters can be found here and a dictionary of output files can be found here.
A report evaluating the efficacy of various configurations of digital contract-tracing mobile phone apps can be found here and the parameters used in the report are documented here. The model was developed by the Pathogen Dynamics group, at the Big Data Institute at the University of Oxford, in conjunction with IBM UK and Faculty. More details about our work can be found at www.coronavirus-fraser-group.org . We suggest running from the latest commit in the master branch or from the latest release tag, which are created at major change points.

Economics Model

adaptER-covid19, and economics model, is attached to the main OpenABM-Covid19 model so the economic effect of Covid-19 can be modelled jointly with the spread of the disease. More information is here.

Compilation

OpenABM-Covid19 requires a C compiler (such as gcc) and the GSL libraries installed. Python installation requires Python 3.7+

cd OpenABM-Covid19/
make all

To install the Python interface, first install SWIG, then:

make install

For developers, the following installs the Python interface inplace, so modifications to the code are applied without needing to reinstall

make dev

Usage

cd OpenABM-Covid19/src
./covid19ibm.exe <input_param_file> <param_line_number> <output_file_dir> <household_demographics_file>

where:

• input_param_file : is a csv file of parameter values (see params.h and the parameter dictionary for further details of the parameters)
• param_line_number : the line number of the parameter file for which to use for the simulation
• output_file_dir : path to output directory (this directory must already exist)
• household_demographics_file : a csv file from which samples are taken to represent household demographics in the model

We recommend running the model via the Python interface (see Examples section with scripts and notebooks below). Alternatively

from COVID19.model import Model, Parameters
import COVID19.simulation as simulation

params = Parameters(
    input_param_file="./tests/data/baseline_parameters.csv",
    param_line_number=1,
    output_file_dir="./data_test",
    input_households="./tests/data/baseline_household_demographics.csv"
)
params.set_param( "n_total", 10000)

model = simulation.COVID19IBM(model = Model(params))
sim   = simulation.Simulation(env = model, end_time = 10 )
sim.steps( 10 )
print( sim.results )     

Examples

The examples/ directory contains some very simple Python scripts and Jupyter notebooks for running the model. The examples must be run from the example directory. In particular

1. examples/example_101.py - the simplest Python script for running the model
2. examples/example_101.ipynb - the simplest notebook of running the model and plotting some output
3. examples/example_102.ipynb - introduces a lock-down based upon the number of infected people and then after the lock-down turns on digital contact-tracing
4. examples/example_extended_output.ipynb - a detailed notebook analysing many aspect of the model and infection transmission.
5. examples/multi_run_simulator.py - an example of running the model multi-threaded

---

Contributing

See document on contributing for further guidelines on contributing features and code to OpenABM-Covid19.

Tests

A full description of the tests run on the model can be found here. Tests are written using pytest and can be run from the main project directory by calling pytest. Tests require Python 3.7 or later. Individual tests can be run using, for instance, pytest tests/test_ibm.py::TestClass::test_hospitalised_zero. Tests have been run against modules listed in tests/requirements.txt in case they are to be run within a virtual environment.

R Packaging

Note: Due to a bug in SWIG #1901, SWIG 4 is unsupported. You'll need to install version 3.0.12.

POSIX platform and Windows require different setups for building the R package.

POSIX setup

On MacOS or Linux systems you will then need to generate and run a configure file (this will need installation of autoconf such as using Homebrew on MacOS, brew install autoconf).

autoconf
./configure

This will generate a file src/Makevars (ignored by git) which includes the necessary GSL compiler and linker flags that R needs when building the C source.

Windows setup

• Set the environment variable LIB_GSL to install prefix of GSL. (It's recommended to compile and install GSL with CMake instead of autotools).
• Install Rtools and add C:\Rtools\usr\bin to your PATH environment variable.
• Install swig 3 using Cygwin or MSYS2 and add swig.exe to your PATH environment variable or set SWIG3.

You do not need to generate and run a configure script. R will use src/Makevar.win which uses the environment variable LIB_GSL to find GSL.

All platforms

To build the R package, you first need to generate the SWIG source files:

make Rswig

If you have multiple versions of SWIG installed, you can set the SWIG3 environment variable to the SWIG 3 executable. Example:

SWIG3=$(HOME)/swig-3.0.12/bin/swig make Rswig

If this is successful, the following files will be generated (ignored by git):

• R/OpenABMCovid19.R
• src/covid19_wrap_R.c

After that, YOU can then open the OpenABM-Covid19.Rproj file in RStudio and build as normal. Remember to re-run make Rswig every time about you modify the C source and/or SWIG interface.

You'll need to install some R packages. In the R console, run:

> install.packages(c("R6", "roxygen2", "devtools", "xptr"))

R console cheat-sheet:

> devtools::document() # Generate the R documentation files
> devtools::build()    # Bundle OpenABMCovid19 into a source package tarball
> devtools::install()  # Install the OpenABMCovid19 package.
> devtools::test()     # Run tests.
> devtools::check()    # Check R package.