birta (version 1.16.0)

birta: Main interface for Bayesian Inference of Regulation of Transcriptional Activity.

Description

The function estimates parameterization of the model and then executes MCMC sampling to infer activity states.

Usage

birta(dat.mRNA, dat.miRNA, TFexpr, limmamRNA=NULL, limmamiRNA=NULL, limmaTF=NULL, nrep=NULL, fdr.mRNA=0.05, fdr.miRNA=0.05, lfc.mRNA=0, lfc.miRNA=0, genesets=NULL, lambda=NULL, sample.weights=TRUE, one.regulator.weight=TRUE, theta_TF=NULL, theta_miRNA=NULL, model=c("all-plug-in", "no-plug-in"), niter=500000, nburnin=100000, thin=50, potential_swaps=NULL, run.pretest=FALSE, condition.specific.inference=TRUE, only_switches=FALSE, weightSampleMean=0, weightSampleVariance=0.01)

Arguments

dat.mRNA

mRNA expression data (ExpressionSet or matrix). IMPORTANT: Replicates must be ordered according to nrep!

dat.miRNA

miRNA expression data (ExpressionSet or matrix).

TFexpr

TF expression data (ExpressionSet or matrix).

limmamRNA

Output of limma analysis for mRNA data (list: pvalue.tab, lm.fit).

limmamiRNA

Output of limma analysis for miRNA data (list: pvalue.tab, lm.fit).

limmaTF

Output of limma analysis for TF expression data (list: pvalue.tab, lm.fit).

nrep

Vector containing four integers. Entry 1 and 2 are the number of miRNA measurement replicates for condition 1 and 2. Entry 3 and 4 are the number of mRNA measurement replicates for condition 1 and 2.

fdr.mRNA

False discovery rate (FDR) cutoff for significance of the logFC for mRNA data.

fdr.miRNA

False discovery rate (FDR) cutoff for significance of the logFC for miRNA data.

lfc.mRNA

Additional logFC cutoff for significance in mRNA data.

lfc.miRNA

Additional logFC cutoff for significance in miRNA data.

genesets

Combined TF / miRNA network. IMPORTANT: Names of TF entries must start with V\$.

lambda

Regularization parameter for edge weights.

sample.weights

Should edge weights be adapted during sampling?

one.regulator.weight

Sould weights of all edges for a regulator to be the same?

theta_TF

Expected fraction of active TFs.

theta_miRNA

Expected fraction of active miRNAs.

model

Type of model. One out of c("all-plug-in", "weight-plug-in", "no-plug-in").

niter

Number of MCMC iterations (AFTER burnin).

nburnin

Number of MCMC iterations UNTIL burnin is assumed to be finished.

potential_swaps

Pre-computed potential swaps (OPTIONAL, see get_potential_swaps).

run.pretest

Initialize miRNA and TF states via the result of a hypergeometric test in order to improve convergence (should be taken with care; advise: only use it in case of observed convergence problems!).

condition.specific.inference

Should inference on TF / miRNA activities be made only RELATIVE to a reference condition or independently in both conditions?

thin

Thinning of Markov chain: only use every thin's sample for posterior computation.

only_switches

Should only switches be performed?

weightSampleMean

Mean for normal distribution used for sampling the omegas.

weightSampleVariance

Variance for normal distribution used for sampling the omegas.

Value

The function returns a list containing the following entries:

miRNAstates1

Probability of each miRNA to be active in condition 1 (only for condition specific sampling).

miRNAstates2

Probability of each miRNA to be active in condition 2 (only for condition specific sampling).

miRNAActivitySwitch

Probability of each miRNA switching its activities (non-condition specific sampling).

TFstates1

Probability of each TF to be active in condition 1 (only for condition specific sampling).

TFstates2

Probability of each TF to be active in condition 2 (only for condition specific sampling).

miRNAactivitySwitch

Probability of each TF switching its activities (non-condition specific sampling).

log_lik_trace

Log-likelihood trace of MCMC sampling.

TFomega

Weights of TF-target graph. (effect of a TF on its targets)

miRNAomega

Weights of miRNA-target graph. (effect of a miRNA on its targets)

genesetsTF

TF-target network. This might be different from the network submitted to the function, due to incompatibilities of network and experimental measurements. Check your warnings and command line output!

genesetsmiRNA

miRNA-target network. This might be different from the network submitted to the function, due to incompatibilities of network and experimental measurements. Check your warnings and command line output!

mRNAexpr

mRNA expression data. This might be different from the matrix submitted to the function, due to incompatibilities of network and experimental measurements. Check your warnings and command line output!

miRNAexpr

miRNA expression data. This might be different from the matrix submitted to the function, due to incompatibilities of network and experimental measurements. Check your warnings and command line output!

TFexpr

TF expression data (only if they are specifically included). This might be different from the matrix submitted to the function, due to incompatibilities of network and experimental measurements. Check your warnings and command line output!

Examples

data(humanSim)
design = model.matrix(~0+factor(c(rep("control", 5), rep("treated", 5))))
colnames(design) = c("control", "treated")
contrasts = "treated - control"
limmamRNA = limmaAnalysis(sim$dat.mRNA, design, contrasts)
limmamiRNA = limmaAnalysis(sim$dat.miRNA, design, contrasts)
sim_result = birta(sim$dat.mRNA, sim$dat.miRNA, limmamRNA=limmamRNA, 
 limmamiRNA=limmamiRNA, nrep=c(5,5,5,5), genesets=genesets, 
 model="all-plug-in", niter=50000, nburnin=10000, 
 sample.weights=FALSE, potential_swaps=potential_swaps)