In vivekkohar/test: Systems biology tool to simulate gene regulatory circuits

knitr::opts_chunk$set(echo = TRUE)

Setting the circuit

This article shows how single model can be simulated using sRACIPE.

suppressWarnings(suppressPackageStartupMessages(library(sRACIPE)))
suppressWarnings(suppressPackageStartupMessages(library(ggplot2)))

racipe <- RacipeSE() # Construct an empty RacipeSE object
data("EMT1") # Load a sample circuit
sracipeCircuit(racipe) <- EMT1
racipe

Generate a random parameter set

One can initialize the racipe object with experimentally determined parameters or generate a random parameter set and then modify the some or all of the parameters. As we are interested in the trajectories of a single model, we set timeSeries to TRUE.

racipe <- sracipeSimulate(racipe, timeSeries = TRUE, plots = FALSE, genIC = TRUE, 
                         genParams = TRUE, integrate = FALSE
                       )

racipe

Modify the parameters or initial conditions

Notice that the colData changed now and contains the randomly generated parameters and initial conditions. These can be accessed and modified using sracipeParams and sracipeIC accessors.

parameters <- sracipeParams(racipe) # Get the parameters
dim(parameters) 
parameters["G_Zeb1"]
parameters[,2]
parameters["G_Zeb1"] <- 10*parameters["G_Zeb1"] # modify parameters
parameters[,2] <- 50 # modify parameters
sracipeParams(racipe) <- parameters # change the parameters to modified values
racipe$G_Zeb1


ic <- sracipeIC(racipe) # Get the initial conditions
dim(ic) 
ic[1]

ic[1] <- 10 # modify initial condition
sracipeIC(racipe) <- ic # change the ic to modified values

Simulate with modified values

racipe <- sracipeSimulate(racipe, timeSeries = TRUE, plots = FALSE, genIC = FALSE, 
                         genParams = FALSE, integrate = TRUE,
                        simulationTime  = 20,
                       integrateStepSize = 0.05
                       )
# Function to plot data
sracipePlotTS <- function(plotData, ...){
  plotData <- t(plotData)
  sexprs <- stack(as.data.frame(plotData))
  colnames(sexprs) <- c("geneExp", "Gene")
  sexprs$time <- rep(as.numeric(rownames(plotData)), ncol(plotData))
  theme_set(theme_bw(base_size = 10))
  ggplot2::qplot(time, geneExp, data = sexprs, group = Gene, colour = Gene,
      geom = "line", ylab = "Gene Expression", xlab = "time" )
}
plotData <- sracipeGetTS(racipe)
p <- sracipePlotTS(plotData)
p + scale_y_log10() # Use log scale

Parameter perturbation

The role of parameter perturbation can be studied by slowly changing one parameter while keeping the other paramters constant. Here we will simulate a large number of models with same parameters except for the parameter being perturbed. We start with a random initial condition each time to capture multiple attractors incase the system is multistable.

selectedParameter <- "G_Zeb1" # parameter to be perturbed
parMin <- 1 # Minimum value of parameter
parMax <- 100 # Maximum value of parameter
nModels <- 30 # Number of models. Parameter value will be uniformly sampled 
# nModels times from parMin to parMax.

parameters <- sracipeParams(racipe)
newParameters <- parameters[rep(seq_len(nrow(parameters)), nModels),]
tmpValue <- seq(from = as.numeric(parMin), 
                  to = as.numeric(parMax),
                  length.out = nModels)
newParameters[selectedParameter] <- tmpValue

circuit <- sracipeCircuit(racipe)
rs <- sracipeSimulate(circuit, genIC = TRUE, genParams = TRUE, integrate = FALSE,
                          numModels = nModels)
sracipeParams(rs) <- newParameters # Modify parameters etc

rs <- sracipeSimulate(rs, genIC = FALSE, genParams = FALSE, integrate = TRUE,
                          numModels = nModels, simulationTime = 100)

# Plot the steady state values
library(ggplot2)
sexprs <- assay(rs,1)
sexprs <- reshape2::melt(t(sexprs))
colnames(sexprs) <- c("bifurParameter","Gene","geneExp")
modParameter <- rep(tmpValue,times = dim(rs)[1])
sexprs$modParameter <- modParameter
theme_set(theme_bw(base_size = 10))
p <- ggplot2::ggplot(sexprs) +
        geom_point(aes(x=modParameter,y=geneExp,color = Gene)) +
        xlab((selectedParameter)) +
        ylab("Gene Expression") 
p + scale_y_log10() # Use log scale