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/src
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

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

To build the R package, you first need to generate the SWIG source files. Do to this, you should use the Makefile at the root of the source directory. You must have the swig command included in your PATH environment variable. To generate the SWIG source files, run:

make Rswig

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

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

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.

Alternatively, you can also build using the command-line instead of RStudio:

• make Rbuild: builds the R source package
• make Rinstall: builds the R binary package
• make Rcheck: check R package for errors

Requirements for building on Windows

If you're building on Windows, there are additional tools that are needed to build the binary package. This is the recommended setup:

• Download precompiled 32 and 64 bits libraries of GSL (installed in C:\gsl).
• Install Rtools (in C:\Rtools).
• Install make & SWIG via Cygwin64 (to run make Rswig).