tni.conditional: Modulators of transcription factor (TF) activity assessed by conditional mutual information analysis.

Description

This function takes a TNI object and a list of candidate modulators, and computes the conditional mutual information over the TF-target interactions in a transcriptional network (with multiple hypothesis testing corrections). For each TF, the method measures the change in the mutual information between the TF and its targets conditioned to the gene expression of a modulator.

Usage

tni.conditional(object, modulators=NULL, tfs=NULL, sampling=35, pValueCutoff=0.01, 
pAdjustMethod="bonferroni", minRegulonSize=15, minIntersectSize=5, 
miThreshold="md", prob=0.99, pwtransform=FALSE, medianEffect=FALSE, 
verbose=TRUE, ...)

Arguments

object

a processed object of class 'TNI' TNI-class evaluated by the methods tni.permutation, tni.bootstrap and tni.dpi.filter.

modulators

a vector with gene identifiers for those considered as candidate modulators. If NULL, the function will assess all TFs in the network, provided that the candidate passes all filtering steps.

tfs

a vector with TF identifiers. If NULL, the function will assess all TFs in the network.

sampling

a single integer value specifying the percentage of the available samples that should be included in the analysis. For example, for each TF-target interaction of a given hub, 'sampling = 35' means that the conditional mutual information will be computed from the top and bottom 35% of the samples ranked by the gene expression of a given candidate modulator.

pValueCutoff

a single numeric value specifying the cutoff for p-values considered significant.

pAdjustMethod

a single character value specifying the p-value adjustment method to be used (see 'p.adjust' for details).

minRegulonSize

a single integer or numeric value specifying the minimum number of elements in a regulon. Gene sets with fewer than this number are removed from the analysis.

minIntersectSize

a single integer or numeric value specifying the minimum number of observed modulated elements in a regulon (as percentage value).

miThreshold

a single character value specifying the underlying distribution used to estimate the mutual information threshold. Options: 'md' and 'md.tf'. In the 1st case, 'miThreshold' is estimated from a pooled null distribution representing random modulators, while in the 2nd case a specific mutual information threshold is estimated for each TF conditioned on the random modulators. In the two options the 'miThreshold' is estimated by permutation analysis (see 'prob'). Alternatively, users can either provide a custom mutual information threshold or a numeric vector with lower (a) and upper (b) bounds for the differential mutual information analysis (e.g. 'c(a,b)').

prob

a probability value in [0,1] used to estimate the 'miThreshold' based on the underlying quantile distribution.

pwtransform

a single logical value specifying whether a Box-Cox power transformation should be used to improve data symmetry (this procedure might be carefully used as a complementary test to check the validity of the inferred associations).

medianEffect

a single logical value specifying whether to assess the median effect of each modulator. This global statistics does not affect the inferential process over single TF-target interactions. This method is still experimental, it can be used as a complementary analysis to chek the overall modulation effect onto all targets listed in a given regulon (this step may require substantial computation time).

verbose

a single logical value specifying to display detailed messages (when verbose=TRUE) or not (when verbose=FALSE)

...

additional arguments passed to tna.graph function.

Value

a data frame in the slot "results", see 'cdt' option in tni.get.

References

Wang, K. et al. Genome-wide identification of post-translational modulators of transcription factor activity in human B cells. Nat Biotechnol, 27(9):829-39, 2009.

Castro, M.A.A. et al. RTN: Reconstruction and Analysis of Transcriptional Networks. Journal Paper (in preparation), 2012.

Examples

Run this code


data(dt4rtn)

# a few TFs for quick demonstration!
tfs4test<-dt4rtn$tfs[c("PTTG1","E2F2","FOXM1","E2F3","RUNX2")]

# create a new TNI object
rtni <- new("TNI", gexp=dt4rtn$gexp, transcriptionFactors=tfs4test)

## Not run: 
# 
# # preprocessing
# rtni <- tni.preprocess(rtni,gexpIDs=dt4rtn$gexpIDs)
# 
# # permutation analysis (infers the reference/relevance network)
# rtni<-tni.permutation(rtni)
# 
# # dpi filter (infers the transcriptional network)
# rtni<-tni.dpi.filter(rtni)
# 
# # ..and a few candidate modulators for demonstration!
# mod4test<-rownames(rtni@gexp)[sample(1:nrow(rtni@gexp),200)]
# 
# # conditional analysis
# rtni<-tni.conditional(rtni, modulators=mod4test, pValueCutoff=1e-3)
# 
# #get results
# cdt<-tni.get(rtni,what="cdt")
# 
# #get summary on a graph object
# g<-tni.graph(rtni,gtype="mmap")
# 
# ###---------------------------------------------
# ### optional: plot the igraph object using RedeR
# library(RedeR)
# 
# #--load reder interface
# rdp<-RedPort()
# calld(rdp)
# 
# #---add graph and legends
# addGraph(rdp,g)
# addLegend.shape(rdp,g)
# addLegend.size(rdp,g)
# addLegend.color(rdp,g,type="edge")
# relax(rdp,p1=50,p5=20)
# 
# ## End(Not run)