DCGL (version 2.1.2)

DRrank: Ranking Regulators by Target Enrichment Density (TED) and Targets' DCL Density (TDD)

Description

The algorithm to rank candidate regulators

Usage

DRrank( DCGs, DCLs, 
	tf2target, expGenes,
	rank.method=c('TED','TDD')[1],
	Nperm=0 )

Arguments

DCGs

a data frame or matrix for DCGs list.

DCLs

a data frame or matrix for DCLs list.

tf2target

a data frame or matrix for TF-to-target interaction pairs.

expGenes

a list for measured genes by array

rank.method

a character string indicating which ranking method to be utilized. The default is 'TED'.

Nperm

permutation times. If Nperm>0, the permutation step will be implemented for TED and TDD methods. The default value for Nperm is 0.

Value

A matrix to display TED or TDD scores and ranks.

Details

DRrank is implemented for ranking potential TFs in terms of their relevance to the phenotypic change or biophysical process of interest. It contains two methods: TED, and TDD.

TED, short for 'Target Enrichment Density', employs Binomial Probability model to quantify the enrichment of a TF's targets in the DCG set, and as such to evaluate which regulators are more likely to be subject-relevant or even causal. Suppose we sift K DCGs from expression profile which contains N genes. If TFi has Ti targets in regulation knowledge, there should be Ti * K / N DCGs appeared in TFi targets list randomly. Actually, it is found that TI DCGs are included in TFi's targets list. The larger TI than Ti * K / N is, the more targets of TFi enriched, the more likely TFi is a relevant or causative regulator.

TDD, short for 'Targets' DCL Density', uses Clustering Coefficient to quantify the density of DCLs among a regulator's targets, and so to judge the importance of a TF. Suppose that TFi has n targets, and that there are k DCLs among these targets. A larger k means more DCLs are bridged by the common TFi. We intuitively assume that, if a TF bridged more TF_bridged_DCL it is of more importance (even if the regulator is not a DCG).

Examples

data(exprs)
## divide exprs into two parts corresponding to condition 1 
## (exprs.1) and condition 2 (exprs.2) respectively
expGenes<-rownames(exprs)
exprs<-exprs[1:100,]
exprs.1<-exprs[1:100,1:16]
exprs.2<-exprs[1:100,17:63]

data(tf2target)
DCp.res<-DCp(exprs.1,exprs.2,
	link.method = 'qth',cutoff=0.25)
DCe.res<-DCe(exprs.1,exprs.2,
	link.method = 'qth',cutoff=0.25,nbins=10,p=0.1)
DCsum.res<-DCsum(DCp.res,DCe.res,DCpcutoff=0.25,DCecutoff=0.4)

## rank all the potential TFs
data(tf2target)
DRrank.TED.res<-DRrank(DCsum.res$DCGs, DCsum.res$DCLs, 
	tf2target, expGenes,
	rank.method=c('TED','TDD')[1],
	Nperm=0)

DRrank.TED.res[1:3,]

DRrank.TDD.res<-DRrank(DCsum.res$DCGs, DCsum.res$DCLs, 
	tf2target, expGenes,
	rank.method=c('TED','TDD')[2],
	Nperm=0)

DRrank.TDD.res[1:3,]