ARTIVA (version 1.2.3)

CP.postDist: Function to compute the CPs posterior distribution for the ARTIVA network model from the the ouput samples of function ARTIVAsubnet.

Description

Using the ouput RJ-MCMC samples of functions ARTIVAsubnet, this function estimates posterior distributions for the number of CPs and their position.

Usage

CP.postDist(CPsamples, burn_in=NULL, segMinLength=2)

Arguments

CPsamples

A matrix with the different iterations (in row) performed with the ARTIVAsubnet function and in column the identified positions for CPs.

burn_in

Number of initial iterations that are discarded for the estimation of the model distribution (posterior distribution). The ARTIVAsubnet function is a RJ-MCMC algorithm which, at each iteration, randomly samples a new configuration of the time-varying regulatory network from probability distributions based on constructing a Markov chain that has the network model distribution as its equilibrium distribution (The equilibrium distribution is obtained when the Markov Chain converges, which requires a large number of iterations). Typically, initial iterations are notconfident because the Markov Chain has not stabilized. The burn-in samples allow to not consider these initial iterations in the final analysis (optional, default: burn_in=NULL, if burn_in=NULL then the first 25% of the iterations is left for burn_in).

segMinLength

Minimal length (number of time points) to define a temporal segment. Must be - strictly - greater than 1 if there is no repeated measurements for each time point in arguments targetData and parentData (optional, default: segMinLength=2).

Value

A list of 4 elements:1) CPnumber: a table containing the approximate posterior distribution for the number of CPs.2) CPposition: a table containing the approximate posterior distribution for the CPs position.3) estimatedCPnumber: number of CP position with the greatest posterior probability according to the approximate posterior distribution for the number of CPs CPnumber.4) estimatedCPpos: a table containing the estimatedCPnumber most significant CP positions according to CPnumber, CPposition and segMinLength (if parameter dyn=1, first CP is 2 and final CP is n+1, where n is the number of time points).

References

Statistical inference of the time-varying structure of gene-regulation networks S. Lebre, J. Becq, F. Devaux, M. P. H. Stumpf, G. Lelandais, BMC Systems Biology, 4:130, 2010.

See Also

ARTIVAsubnet, ARTIVAnet, plotCP.postDist, ARTIVAsubnetAnalysis

Examples

# Load the ARTIVA R package
library(ARTIVA)

# Load the dataset with simulated gene expression profiles
data(simulatedProfiles)

# Name of the target gene to be analyzed with ARTIVA 
targetGene = 1

# Names of the parent genes (typically transcription factors) 
parentGenes = c("TF1", "TF2", "TF3", "TF4", "TF5")

# run ARTIVAsubnet

# Note that the number of iterations in the RJ-MCMC sampling is reduced 
# to 'niter=20000' in this example, but it should be increased (e.g. up to
# 50000) for a better estimation.
## Not run: 
# ARTIVAtest = ARTIVAsubnet(targetData = simulatedProfiles[targetGene,],
#   parentData = simulatedProfiles[parentGenes,],
#   targetName = targetGene,
#   parentNames = parentGenes,
#   segMinLength = 2,
#   edgesThreshold = 0.6, 
#   niter= 20000,
#   savePictures=FALSE)
# 
# # compute the PC posterior distribution with other parameters
# outCPpostDist = CP.postDist(ARTIVAtest$Samples$CP, burn_in=10000, 
# 			    segMinLength=3)
# 
# # plot the CP posterior distribution
# plotCP.postDist(outCPpostDist, targetName=paste("Target", targetGene), 
# 		  estimatedCPpos=outCPpostDist$estimatedCPpos)
# ## End(Not run)