The goal of ModReg is to infer whether a modulator activates or inhibits the activity of a regulator. The hypothesis here is that if a modulator can control the function of a regulator, the modulator’s expression level would affect the correlation between the regulator and its targets. For example, chromatin factors can modulate the chromatin architecture to affect the function of transcription factors in response to intrinsic and extrinsic signalling cues. In our study, we found that SIRT7, a type of histone deacetylase, is able to activate the activity of transcription factor NFE2L2 in liver cancer because high levels of SIRT7 expression can result in stronger correlations between NFE2L2 and its targets, as shown in the tutorial below.
You can install the development version of ModReg from
GitHub with:
# install.packages("devtools")
devtools::install_github("beibeiru/ModReg")This example shows how to infer the modulation relation between SIRT7
and NFE2L2 using the TCGA liver cancer dataset. Also, two visualization
functions are designed to present the output of ModReg package.
library(ModReg)
# load example data
load(file.path(system.file(package = "ModReg"), "extdata/exampleData.rda"))
# show loaded data
ls()
## [1] "exprMatrix" "modulator" "regulator" "regulon" "target"
# TCGA LIHC RNA-Seq data
head(exprMatrix[,1:3])
## TCGA.2V.A95S.01 TCGA.2Y.A9GS.01 TCGA.2Y.A9GT.01
## A1BG 9.507621 9.4190609 11.30630809
## A1BG-AS1 2.619266 1.8605297 3.28726563
## A1CF 4.252538 5.5715862 5.38414905
## A2M 13.191258 9.8801464 9.31354503
## A2M-AS1 3.954040 0.1739792 0.05847934
## A4GALT 1.394355 2.3000737 1.95371228
# Regulon (target genes) of NFE2L2
head(regulon)
## [1] "RENBP" "RGSL1" "MMADHC" "RNF13" "RNFT1" "RNFT2"
# run ModReg
ModReg_obj <- ModReg(exprMatrix, modulator="SIRT7", regulator="NFE2L2", regulon)
# show the output of ModReg
str(ModReg_obj)
## List of 5
## $ modulator.regulator :'data.frame': 1 obs. of 8 variables:
## ..$ modulator : chr "SIRT7"
## ..$ mode : chr "activate"
## ..$ regulator : chr "NFE2L2"
## ..$ signif_ratio : num 0.643
## ..$ activate_ratio: num 0.643
## ..$ inhibit_ratio : num 0
## ..$ dami : num 0.129
## ..$ dami_p : num 0
## $ modulator.regulator.target: num [1:168, 1:7] -0.046 0.1553 0.1893 0.0936 0.4036 ...
## ..- attr(*, "dimnames")=List of 2
## .. ..$ : chr [1:168] "IRF2" "RBL2" "TBC1D23" "CCDC121" ...
## .. ..$ : chr [1:7] "modulator_low_r" "modulator_low_p" "modulator_high_r" "modulator_high_p" ...
## $ regulon : chr [1:168] "RENBP" "RGSL1" "MMADHC" "RNF13" ...
## $ cutoff.modulator : num 0.25
## $ permut_num : num 1000
# Modulation relation
ModReg_obj$modulator.regulator
## modulator mode regulator signif_ratio activate_ratio inhibit_ratio dami dami_p
## Value SIRT7 activate NFE2L2 0.6428571 0.6428571 0 0.1294076 0
# The correlation between NFE2L2 and its regulon under the low and high level of SIRT7 expression.
head(ModReg_obj$modulator.regulator.target)
## modulator_low_r modulator_low_p modulator_high_r modulator_high_p diff_z diff_p diff_padj
## IRF2 -0.04599914 6.615081e-01 0.7236200 2.536316e-16 6.448156 1.132194e-10 1.902086e-08
## RBL2 0.15532064 1.371105e-01 0.7886838 6.255371e-21 6.113533 9.744920e-10 8.185733e-08
## TBC1D23 0.18925325 6.923779e-02 0.7795143 3.459080e-20 5.719207 1.070227e-08 5.993269e-07
## CCDC121 0.09356243 3.723719e-01 0.7128720 1.098306e-15 5.360945 8.278766e-08 3.444419e-06
## SEPT10 0.40361910 6.026098e-05 0.8400530 6.694026e-26 5.322213 1.025125e-07 3.444419e-06
## LIMD2 -0.08778239 4.027554e-01 -0.6987483 6.828116e-15 -5.211212 1.876114e-07 5.253118e-06
The output shows that SIRT7 activates transcriptional regulation of
NFE2L2. plotModReg can be used to visualize this modulation relation.
The left and right panels present the patient group with low (M−) and
high (M+) level of SIRT7 expression, in which rows are targets and
columns are patient samples, respectively. In each patient group (M− or
M+), the samples are sorted based on NFE2L2 expression level.
plotModReg(exprMatrix,ModReg_obj)
In current example, 64.3% transcriptional targets of NFE2L2 could be
activated by high level of SIRT7 expression (adjusted p<0.05).
plotModRegTar can be used to visualize any tripartite relation of
interest. For example, NFE2L2 has significant different correlations
with its target IRF2 dependent on SIRT7 expression level.
plotModRegTar(exprMatrix, modulator="SIRT7", regulator="NFE2L2", target="IRF2")
ModReg is comprised of two steps. At first, all tripartite relations of Modulator-Regulator-Target were estimated based on differential correlation analysis. Specifically, all tumor samples were separated into the top (M+) and bottom (M−) 25% of samples in which the modulator was most and least expressed. In these two sets (M+ and M−), the significance of differential correlation relationships between one regulator and its targets were quantified using the difference in z-scores by Fisher z-transformation. A two-sided p value can be calculated using the standard normal distribution, and Benjamini–Hochberg method was used to adjust P-values for multiple hypothesis tests. Significant tripartite relations were kept for the next step (adjusted p<0.05).
Secondly, ModReg calculated the differential average mutual information (DAMI) of the significant regulated targets from step 1, which equals to the difference of average mutual information (i.e., dependence relationship) of all significant regulated targets between M+ and M− sample sets. Thus, DAMI quantified how much the modulator was able to affect the fucntion of the regulator (DAMI > 0 for activation and DAMI < 0 for inhibition, respectively). Permutation test is used to estimate the empirical distribution of DAMI by randomly selecting a target set of equal size with the significant regulated targets from the complete regulon of the regulator, and the associated p value was estimated as the percentage of random trials with an absolute value of DAMI greater than the absolute value of the measured DAMI. Based on the DAMI and p value here, ModReg can determine the modulation mode of the modulator and the regulator.
Beibei Ru, Jianlong Sun, Qingzheng Kang, Yin Tong, Jiangwen Zhang. A framework for identifying dysregulated chromatin regulators as master regulators in human cancer. Bioinformatics. 2019; 35(11):1805-1812. [Link]