# Individual-based SARS-CoV-2 transmission model, practical 2
library(ggplot2)
## Model parameters
<- 0.5 # Transmission parameter
beta <- 1e-5 # Importation rate
iota <- 1/2.5 # Rate of transitioning out of latent state
delta <- 1/5 # Rate of transitioning out of infectious state
gamma <- 1/180 # Rate of waning immunity
omega
<- 1 # Time step of simulation (1 day)
dt <- 365*2 # Duration of simulation (2 years)
days <- days/dt # Total number of time steps
steps <- 1000 # Population size
n
## Some helper functions
# Calculates infectiousness as a function of state and age: zero if state is
# not "I"; nonzero if state is "I", and slightly decreasing with age
<- function(state, age) {
infectiousness ifelse(state == "I", 1.25 - age / 160, 0)
}
# Calculates susceptibility of individuals with antibody level(s) ab
<- function(ab) {
susceptibility pnorm(ab, 5, 1, lower.tail = FALSE)
}
# Generates n random delays from the latent-period distribution
# (approximately 2 days, on average)
<- function(n) {
latent_delay rlnorm(n, meanlog = 0.5, sdlog = 0.6)
}
# Generates n random delays from the infectious-period distribution
# (approximately 5 days, on average)
<- function(n) {
infectious_delay rlnorm(n, meanlog = 1.5, sdlog = 0.5)
}
# Generates n random increments to antibody levels following recovery
<- function(n) {
ab_increment rnorm(n, mean = 12, sd = 2)
}
## Data frame to store simulation results
<- data.frame(ts = 1:steps, S = 0, E = 0, I = 0, R = 0)
results
## Initialize simulation
# Set the seed for the pseudorandom number generator, for reproducibility
set.seed(12345)
# Initialize state variables
<- rep("S", n) # Each individual's state: start with all susceptible
state 1:10] <- "E" # Start 10 individuals in the "exposed" state
state[
## Run simulation
# Initialize progress bar
<- txtProgressBar(min = 1, max = steps, style = 3)
bar
# Loop over each time step . . .
for (ts in 1:steps) {
# Calculate the force of infection
<- beta * sum(state == "I") / n + iota
lambda
# Loop through each host . . .
for (i in 1:n) {
# Update individual i's non-state variables
# . . .
# Update individual i's state
if (state[i] == "S") {
# Transition S -> E (infection) at rate lambda
if (runif(1) < 1 - exp(-lambda * dt)) {
<- "E"
state[i]
}else if (state[i] == "E") {
} # Transition E -> I (latent to infectious) at rate delta
if (runif(1) < 1 - exp(-delta * dt)) {
<- "I"
state[i]
}else if (state[i] == "I") {
} # Transition I -> R (infectious to recovered) at rate gamma
if (runif(1) < 1 - exp(-gamma * dt)) {
<- "R"
state[i]
}else if (state[i] == "R") {
} # Transition R -> S (waning of immunity) at rate omega
if (runif(1) < 1 - exp(-omega * dt)) {
<- "S"
state[i]
}
}
}
# Save population state for this time step
"S"] <- sum(state == "S")
results[ts, "E"] <- sum(state == "E")
results[ts, "I"] <- sum(state == "I")
results[ts, "R"] <- sum(state == "R")
results[ts,
# Update progress bar; close progress bar if we are finished
setTxtProgressBar(bar, ts)
if (ts == steps) {
close(bar)
}
}
## Plot simulation results
ggplot(results) +
geom_line(aes(x = ts, y = S, colour = "S")) +
geom_line(aes(x = ts, y = E, colour = "E")) +
geom_line(aes(x = ts, y = I, colour = "I")) +
geom_line(aes(x = ts, y = R, colour = "R"))
08. Stochastic individual-based models (practical 2)
Practical 2. Adding more complex dynamics to the model
For practical 2, start with the following code:
Return to the practical here.