NonLinearNet: Dynamic Bayesian Network Inference Using Non-Linear Interactions

Description

Run Bayesian inference of non-linear interaction network. Non linear interactions are modelled using Penalised Splines. The function generates MCMC chains that can later be analysed.

Usage

NonLinearNet( resultsFolder,        timeSeries,   ParamVec = NULL, 
		 chains = 2, user.seeds = NULL, Regulators = NULL,      
	       fixMe = NULL)

Arguments

resultsFolder

Name of output folder. The folder will be created and the output of the run will be placed there.

timeSeries

Data matrix containing gene expression time series. Where genes will be placed in rows and time points in columns. Gene names may be included as row names.

ParamVec

A parameter vector created using "mcmc.defaultParams_nonLinear". If none is given, default parameters will be used. The vector contains parameters associated to the priors as well as MCMC run length. (See mcmc.defaultParams_nonLinear)

chains

Number of MCMC chains to run.

user.seeds

An optional vector with seeds to use for MCMC chains.

Regulators

An optional vector with the indices of which genes are regulators. If provided, all non-regulator genes will not be allowed to regulate.

fixMe

An optional matrix of size genes x genes, where columns represent regulators and rows regulated genes. The matrix informs the model of network connections known to be present/absent. For each position use either 0 (no regulation, fix off), 1 (known regulatory interaction, fix on) or NaN (no information, do not fix).

Value

For each chain run, a folder (chain1, chain2, ...) will be created and the output of the MCMC run will be placed there. The files will be Gamma_mcmc (the indicator variables of Gibbs variable selection), Lambda_mcmc (the precision of each regression), Mu_mcmc (the intercept of each regression), Rho_mcmc (the network connectivity parameter), Tau_mcmc (the "smoothness parameter"), all_f (posterior mean of all functions), all_f_sqr (posterior mean of the square of all functions) and Full_F_sqr (posterior mean of the square of the sum of all functions, for each regression). For the files all_f and all_f_sqr functions are placed in column-wise order. The file is filled by placing all interactions for each regression one after another.

References

Morrissey, E.R., Juarez, M.A., Denby, K.J. and Burroughs, N.J. 2011 Inferring the time-invariant topology of a nonlinear sparse gene regulatory network using fully Bayesian spline autoregression Biostatistics 2011; doi: 10.1093/biostatistics/kxr009

Examples

Run this code

  # Synthetic data
  data(Athaliana_ODE)
  # Reduced data set to 3 genes 20 TP for faster run
  Athaliana_ODE.reduced <- Athaliana_ODE[c(1,3,5),1:20]
  # Folder where raw runs will be kept and later analysed
  output.folder <- paste(tempdir(), "/ExampleNonLinearNet", sep = "")
  # Run network inference and place raw results in output.folder
  NonLinearNet(output.folder , Athaliana_ODE.reduced)
  # Analyse raw results, place analysis plots and files in output.folder
  analyse.output(output.folder, Athaliana_ODE.reduced)