Simulating and Analysing Transmission Chain Statistics Using Branching Process Models • epichains

epichains is an R package to simulate, analyse, and visualize the size and length of branching processes with a given offspring distribution. These models are often used in infectious disease epidemiology, where the chains represent chains of transmission, and the offspring distribution represents the distribution of secondary infections caused by an infected individual.

epichains re-implements bpmodels by providing bespoke functions and data structures that allow easy manipulation and interoperability with other Epiverse-TRACE packages, for example, superspreading and epiparameter, and potentially some existing packages for handling transmission chains, for example, epicontacts.

epichains is developed at the Centre for the Mathematical Modelling of Infectious Diseases at the London School of Hygiene and Tropical Medicine as part of the Epiverse Initiative.

Installation

The latest development version of the epichains package can be installed via

# check whether {pak} is installed
if (!require("pak")) install.packages("pak")
pak::pak("epiverse-trace/epichains")

To load the package, use

library("epichains")

Quick start

epichains provides three main functions:

simulate_chains(): simulates transmission chains using a simple branching process model that accepts an index number of cases that seed the outbreak, a distribution of offspring per case, and a chain statistic to track (size or length/duration). It optionally accepts other population related inputs such as the population size (defaults to Inf) and percentage of the population initially immune (defaults to 0). This function returns an object with columns that track information on who infected whom, the generation of infection and, if a generation time function is specified, the time of infection.
simulate_summary(): provides a performant version of simulate_chains() that only tracks and return a vector of realized chain sizes or lengths/durations for each index case without details of the infection tree.
likelihood(): calculates the loglikelihood (or likelihood, depending on the value of log) of observing a vector of transmission chain sizes or lengths.

The objects returned by the simulate_*() functions can be summarised with summary(). Running summary() on the output of simulate_chains() will return the same output as simulate_summary() using the same inputs.

Objects returned from simulate_chains() can be aggregated into a <data.frame> of cases per time or generation with the function aggregate(). The aggregated results can also be passed on to plot() with its own arguments to customize the resulting plots.

Simulation

For the simulation functionality, let’s look at a simple example where we simulate a transmission chain with 20 index cases, a constant generation time of 3, and a poisson offspring distribution with mean 1. We are tracking the chain “size” statistic and will cap all chain sizes at 25 cases. We will then look at the summary of the simulation, and aggregate it into cases per generation, and plot it.

set.seed(32)
# Simulate chains
sim_chains <- simulate_chains(
  index_cases = 20,
  statistic = "size",
  offspring_dist = rpois,
  stat_max = 25,
  generation_time = function(n) {rep(3, n)}, # constant generation time of 3
  lambda = 1 # mean of the Poisson distribution
)
# View the head of the simulation
head(sim_chains)
#>    infectee_id sim_id infector_id generation time
#> 21           1      2           1          2    3
#> 22           2      2           1          2    3
#> 23           3      2           1          2    3
#> 24           4      2           1          2    3
#> 25           6      2           1          2    3
#> 26           7      2           1          2    3

# Summarise the simulation
summary(sim_chains)
#> `epichains_summary` object 
#> 
#>  [1]   5  17   4   8   1  16   9 Inf   5  18   5   1 Inf  24   1  14  19   2   4
#> [20]  14
#> 
#>  Simulated tree sizes: 
#> 
#> Max: 24
#> Min: 1

# Aggregate the simulation into cases per generation
chains_agrgegated <- aggregate(sim_chains, by = "generation")

# plot the aggregated results
plot(chains_agrgegated, type = "b")

Inference

Let’s look at the following example where we estimate the log-likelihood of observing a hypothetical chain_lengths dataset.

set.seed(32)
# randomly generate 20 chain lengths between 1 to 40
chain_lengths <- sample(1:40, 20, replace = TRUE)
chain_lengths
#>  [1]  6 11 20  9 40 33 39 27  6 12 39 35  9 25  6 15 12  6 37 35

# estimate loglikelihood of the observed chain sizes
likelihood_eg <- likelihood(
  chains = chain_lengths,
  statistic = "length",
  offspring_dist = rpois,
  lambda = 0.99
)
# Print the estimate
likelihood_eg
#> [1] -104.2917

Each of the listed functionalities is demonstrated in detail in the “Getting Started” vignette.

Package vignettes

The theory behind the models provided here can be found in the theory vignette.

We have also collated a bibliography of branching process applications in epidemiology. These can be found in the literature vignette.

Specific use cases of epichains can be found in the online documentation as package vignettes, under “Articles”.

As far as we know, below are the existing R packages for simulating branching processes and transmission chains.

bpmodels: provides methods for analysing the size and length of transmission chains from branching process models. epichains is intended to supersede {bpmodels}.
ringbp: a branching process model, parameterised to the 2019-nCoV outbreak, and used to quantify the potential effectiveness of contact tracing and isolation of cases.
covidhm: code for simulating COVID-19 dynamics in a range of scenarios across a real-world social network. The model is conceptually based on {ringbp}.
epicontacts: provides methods for handling, analysing, and visualizing transmission chains and contact-tracing data/linelists.
simulist: uses a branching process model to simulate individual-level infectious disease outbreak data, including line lists and contact tracing data. This package is part of the Epiverse-TRACE Initiative.
superspreading: provides a set of functions to estimate and understand individual-level variation in transmission of infectious diseases from data on secondary cases. These are useful for understanding the role of superspreading in the spread of infectious diseases and for informing public health interventions.
earlyR: estimates the reproduction number (R), in the early stages of an outbreak. The model requires a specified serial interval distribution, characterised by the mean and standard deviation of the (Gamma) distribution, and data on daily disease incidence, including only confirmed and probable cases.
projections: uses data on daily incidence, the serial interval (time between onsets of infectors and infectees) and the reproduction number to simulate plausible epidemic trajectories and project future incidence. It relies on a branching process where daily incidence follows a Poisson or a Negative Binomial distribution governed by a force of infection.
simulacr: simulates outbreaks for specified values of reproduction number, incubation period, duration of infectiousness, and optionally reporting delays. Outputs a linelist stored as a data.frame with the class outbreak, including information on transmission chains; the output can be converted to <epicontacts> objects for visualisation.
outbreakr: implements a Bayesian approach for reconstructing outbreak data from pathogen genome sequences. It also implements a tool for outbreak simulation.
outbreakr2: a Bayesian framework for integrating epidemiological and genetic data to reconstruct transmission trees of densely sampled outbreaks. It re-implements, generalises and replaces the model of outbreaker, and uses a modular approach which enables fine customisation of priors, likelihoods and parameter movements.
o2geosocial: integrates geographical and social contact data to reconstruct transmission chains. It combines the age group, location, onset date and genotype of cases to infer their import status, and their likely infector.
nosoi: simulates agent-based transmission chains by taking into account the influence of multiple variables on the transmission process (e.g. dual-host systems (such as arboviruses), within-host viral dynamics, transportation, population structure), alone or taken together, to create complex but relatively intuitive epidemiological simulations.
TransPhylo: reconstructs infectious disease transmission using genomic data.

Reporting bugs

To report a bug please open an issue.

Contribute

We welcome contributions to enhance the package’s functionalities. If you wish to do so, please follow the package contributing guide.

Code of conduct

Please note that the epichains project is released with a Contributor Code of Conduct. By contributing to this project, you agree to abide by its terms.

Citing this package

citation("epichains")
#> To cite package 'epichains' in publications use:
#> 
#>   Azam J, Finger F, Funk S (2024). _epichains: Simulating and Analysing
#>   Transmission Chain Statistics Using Branching Process Models_. R
#>   package version 0.0.0.9999,
#>   https://epiverse-trace.github.io/epichains/,
#>   <https://github.com/epiverse-trace/epichains>.
#> 
#> A BibTeX entry for LaTeX users is
#> 
#>   @Manual{,
#>     title = {epichains: Simulating and Analysing Transmission Chain Statistics Using
#> Branching Process Models},
#>     author = {James M. Azam and Flavio Finger and Sebastian Funk},
#>     year = {2024},
#>     note = {R package version 0.0.0.9999, 
#> https://epiverse-trace.github.io/epichains/},
#>     url = {https://github.com/epiverse-trace/epichains},
#>   }

epichains: Methods for simulating and analysing the size and length of transmission chains from branching process models