| Version: | 1.1.1.1 |
| Title: | Inferring Large-Scale Gene Networks with C3NET |
| Author: | Gokmen Altay, Frank Emmert-Streib |
| Description: | Allows inferring gene regulatory networks with direct physical interactions from microarray expression data using C3NET. |
| Maintainer: | Gokmen Altay <altayscience@gmail.com> |
| License: | GPL (≥ 3) |
| Depends: | R (≥ 2.12.1), igraph |
| Repository: | CRAN |
| Repository/R-Forge/Project: | c3net |
| Repository/R-Forge/Revision: | 23 |
| Date/Publication: | 2022-06-24 12:29:23 UTC |
| Packaged: | 2022-06-24 12:13:53 UTC; hornik |
| NeedsCompilation: | no |
C3NET second step: Selecting maximum valued elements
Description
c3 takes the mutual information matrix as input and implements the second step of C3NET - see details.
Usage
c3( mim, sym = TRUE )Arguments
mim |
A symmetric square mutual information matrix, where the elements (i,j) correspond to the mutual information |
sym |
Decides the output matrix to be symmetric or not. |
Details
The C3NET algorithm consists of two main steps. The first step is the same as for relevance networks (RELNET), where all the non-significant mutual information values in the matrix are eliminated if statistically not significant. The second step of C3NET keeps all maximum valued mutual information values for each row in the matrix and sets the rest of the elements in the matrix zero (the diagonal of the matrix is ignored). The output is normally symmetric matrix but if the argument sym is set to FALSE then the output becomes non-symmetric.
Value
c3 implements the second step of the C3NET algorithm returning a symmetric mutual information matrix. Specifically, the non-zero elements in the returned matrix represent undirected links between variables which are statistically significant (tested in the first step of C3NET).
References
G. Altay, F. Emmert-Streib, "Inferring the conservative causal core of gene regulatory networks", BMC Systems Biology (2010) 4:132.
See Also
makemim, copula, sigtestp, sigtestMTC, netplot, c3net
Examples
data(expdata)
expdata <- copula(expdata)
mim <- makemim(expdata)
Ic <- mean(mim[upper.tri(mim)]) #Example cut-off for the first step of C3NET
# Ic <- 2 can be set for the example.
mim[mim < Ic] <-0 #nonsignificant values eliminated wrt C3NET step 1.
net <- c3(mim) # regulatory network inferred (non zero elements stand for links of
#the predicted network)
All in one function to infer network with C3NET
Description
c3net takes a data set as input and computes the inferred network using C3NET - see details.
Usage
c3net(dataset, cop=TRUE, alpha=0.01, methodstep1="cutoff", cutoffMI= 0,
MTCmethod="BH", itnum=5, network=FALSE)Arguments
dataset |
Data set where rows are variables (e.g. genes) and columns are samples. |
cop |
The data set is copula transformed by default. If not wanted, then "cop" is set to FALSE. |
alpha |
Statistical significance threshold |
itnum |
Number of iterations to resample data to get sampling distribution. |
methodstep1 |
If it is "cutoff" then the input cutoffMI is used as threshold for elimination in step 1. If it is "MTC" then MTCmethod input is used as the specific multiple testing correction method to employ. If it is "justp" then only significance threshold alpha is used to compute significance threshold MI. |
MTCmethod |
Available option are "holm", "hochberg", "hommel", "bonferroni", "BH", "BY". |
cutoffMI |
If methodstep1 is set to "cutoff" then this value is used for MI threshold. If this is set to 0 (or by default) it uses mean MI for threshold. |
network |
If TRUE, the network is plotted in the end. |
Details
For Step 1 of C3NET, there are three options that can be used. If methodstep1="cutoff" then cutoffMI input is taken as the significance threshold for step 1. In case cutoffMI is set to 0 then by default mean MI is taken as cutoffMI. If methodstep1 is set to "MTC" then the method from input MTCmethod, itnum for iteration number to get null distribution and alpha for statistical significance is taken for computations. If methodstep1 is set to "justp" or anything else then just alpha and itnum is used for computations.
Value
c3net returns a symmetric mutual information matrix, which is obtained after implementing C3NET. Specifically, non-zero elements in the returned matrix represents undirected link between variables. The inferred network may also be plotted if the argument network is set TRUE.
References
G. Altay, F. Emmert-Streib, "Inferring the conservative causal core of gene regulatory networks", BMC Systems Biology (2010) 4:132.
See Also
makemim, copula, c3, sigtestp, sigtestMTC
Examples
data(expdata)
data(truenet)
net <- c3net(expdata, network=TRUE)
scores <- checknet(net,truenet)
Validating the inferred network
Description
checknet takes the inferred mutual information matrix and the true network for validation. - see details.
Usage
checknet( finalrelationmatrix, realrelationmatrix)Arguments
finalrelationmatrix |
The inferred symmetric mutual information matrix, where i,j th element is the mutual information |
realrelationmatrix |
A symmetric reference connection matrix, where a 1 at i,j th element defines the connection between variables i and j and non-connection is represented by 0. The diagonal is all zero. |
Value
checknet returns a vector with 6 elements, that contains the parameters as output <- c(precision, Fscore, recall, TP, FP, FN).
References
G. Altay, F. Emmert-Streib, "Inferring the conservative causal core of gene regulatory networks", BMC Systems Biology, (2010) 4:132.
See Also
makemim, copula, c3, c3net, c3, makemim, copula, sigtestp, sigtestMTC
Examples
data(expdata)
data(truenet)
expdata <- copula(expdata)
mim <- makemim(expdata)
Ic <- 2 #Example cut-off value for mutual information for the first step of C3NET
mim[mim < Ic] <-0 #nonsignificant values eliminated wrt C3NET step 1.
net <- c3(mim) # regulatory network inferred (non zero elements stand for links of
# the predicted network)
scores <- checknet(net,truenet)
Copula transformation of data set
Description
copula takes a data set as input and applies a copula transformation - see details.
Usage
copula( expdata )Arguments
expdata |
Data set; rows correspond to variables (e.g. genes) and columns to samples. |
Details
The data set can be copula transformed for more stable estimations of the mutual information matrix. Each row of the data set (gene samples of each gene) is copula transformed and obtained rdata. Here, the formula for conversion is copuladata =(rdata- 0.5)/numberofsamples.
Value
copula returns a copula transformed data set.
References
G. Altay, F. Emmert-Streib, "Inferring the conservative causal core of gene regulatory networks", BMC Systems Biology (2010) 4:132.
See Also
Examples
data(expdata)
expdata <- copula(expdata)
Example data set
Description
expdata This data set has been generated by SynTRen. The source network corresponds to a subnetwork of E.coli (file name "EColi_full_HongWu_Ma_NAR2004.sif" from Ma HW, et. al. Nuc. Ac. Res. 2004).
Usage
data( expdata )Arguments
expdata |
Example data; rows correspond to variables (e.g. genes) and columns to samples. |
Details
The example is also given to provide a sample data set so that one can take it as reference while formatting the new data set as entry to c3net. Data sets must be complete and no empty or non-numerical character allowed.
Value
expdata returns an example data set where rows are variables (e.g. genes) and columns are samples. It consists of 400 genes, 898 edges, and 800 steady-state samples.
References
Ma HW, Kumar B, Ditges U, Gunzer F, Buer J, Zeng AP: An extended transcriptional regulatory network of Escherichia coli and analysis of its hierarchical structure and network motifs. Nucleic Acids Res 2004 , 32:6643-6649
Bulcke, T.V., Leemput, K.V., Naudts, B., et. al. (2006) Syntren: a generator of synthetic gene expression data for design and analysis of structure learning algorithms. BMC Bioinformatics, 7(1), 43.
G. Altay, F. Emmert-Streib, "Inferring the conservative causal core of gene regulatory networks", BMC Systems Biology (2010) 4:132.
See Also
Examples
data(expdata)
Constructing mutual information matrix
Description
makemim takes the data set as input and computes mutual information values for each pair of variables (e.g. gene). - see details.
Usage
makemim( expdata )Arguments
expdata |
Data matrix where rows correspond to variables (e.g. genes) and columns to samples. |
Details
A mutual information matrix is generated from a data set using an empirical Gaussian estimator.
Value
makemim returns a symmetric mutual information matrix using empirical Gaussian estimator.
References
G. A. Darbellay, I. Vajda, "Estimation of the information by an adaptive partitioning of the observation space", IEEE Trans. Inf. Theory vol.45 (1999)
G. Altay, F. Emmert-Streib, "Inferring the conservative causal core of gene regulatory networks", BMC Systems Biology, (2010) 4:132.
See Also
Examples
data(expdata)
expdata <- copula(expdata)
mim <- makemim(expdata)
Plotting the inferred network
Description
netplot takes the inferred mutual information matrix and plots a network wrt nonzero elements of the matrix - see details.
Usage
netplot(gnet)Arguments
gnet |
A symmetric squared mutual information matrix that is inferred as final connection matrix, where i,j th element is the mutual information |
Details
A labelled undirected network is plotted wrt the row (column) names of the input matrix. When plotted, select "Layout" and then "Fruchterman-Reingold" and "OK" to get a better view. To remove labels, select "View" and deselect "Labels". To change the colors or size of nodes and edges, from "Select", click on "all edges" or "vertices" and on the network right click on nodes and change color or size.
Value
netplot returns an undirected network.
References
G. Altay, F. Emmert-Streib, "Inferring the conservative causal core of gene regulatory networks", BMC Systems Biology (2010) 4:132.
CSARDI G, NEPUSZ T.: igraph-package for R, 2008
See Also
makemim, copula, c3, sigtestp, sigtestMTC
Examples
data(expdata)
data(truenet)
expdata <- copula(expdata)
mim <- makemim(expdata)
Ic <- 2 #Example cut-off for the first step of C3NET
mim[mim < Ic] <-0 #nonsignificant values eliminated wrt C3NET step 1.
net <- c3(mim) # regulatory network inferred (non zero elements stand for links of
# the predicted network)
netplot(net)
Elimination of nonsignificant edges by application of a procedure for multiple testing correction
Description
sigtestMTC takes a data set as input and computes the mutual information matrix in which the nonsignificant elements are eliminated by a multiple testing correction - see details.
Usage
sigtestMTC( data, alpha, itnum, methodsig="BH")Arguments
data |
Data set where rows correspond to variables (e.g. genes) and columns to samples. |
alpha |
Significance level |
itnum |
Number of iterations to resample data to estimate sampling distribution. |
methodsig |
A procedure to perform a multiple testing correction, either controling the false discovery rate (FDR) or the family-wise error. Available option are "holm", "hochberg", "hommel", "bonferroni", "BH", "BY". |
Details
The data set is resampled for each iteration and at the end the sampling distribution is obtained.
Value
sigtestMTC returns an environment res that contains the new MI matrix, res$Inew, that is obtained after the elimination of nonsignificant elements with a MTC procedure. For advanced users, who are able to write code in R, we provided the variables obtained in the function so that one might want to make further analysis without running the function again. Please see the source code for the additional return values of the environment res.
References
G. Altay, F. Emmert-Streib, "Inferring the conservative causal core of gene regulatory networks", BMC Systems Biology (2010) 4:132.
See Also
Examples
data(expdata)
data(truenet)
alpha <- 0.001
itnum <-2
res <- sigtestMTC( expdata, alpha, itnum, methodsig="BH")
net <- c3(res$Inew) # regulatory network inferred (non zero elements stand for links of
# the predicted network)
scores <- checknet(net,truenet)
Significance test for elimination of nonsignificant edges
Description
sigtestp takes the data set as input and computes the mutual information matrix in which the nonsignificant elements are eliminated by using a significance threshold only (no MTC) - see details.
Usage
sigtestp( data, alpha, itnum)Arguments
data |
Data set where rows are variables (e.g. genes) and columns are samples. |
alpha |
Statistical significance threshold |
itnum |
Number of iterations to resample data to get sampling distribution. |
Details
The data set is resampled completely at each iteration and at the end the sampling distribution is obtained for using in significance test.
Value
sigtestp returns environment res that contains the obtained threshold value res$I0. For advanced users, who are able to write code in R, we provided the variables obtained in the function so that one might want to make further analysis without running the function again. Please see the source code for the additional return values of the environment res.
References
G. Altay, F. Emmert-Streib, "Inferring the conservative causal core of gene regulatory networks", BMC Systems Biology (2010) 4:132.
See Also
makemim, copula, c3,c3, sigtestMTC
Examples
data(expdata)
data(truenet)
alpha <- 0.001
itnum <- 2
res <- sigtestp( expdata, alpha, itnum)
net <- c3(res$Inew) # regulatory network inferred (non zero elements stand for links of
# the predicted network)
scores <- checknet(net,truenet)
Reference, e.g. true, network of the example data set
Description
truenet A symmetric matrix, where non-zero elements correspond to true connections.
Usage
data( truenet )Arguments
truenet |
Reference (true) network of the example data set. |
Details
A symmetric matrix, where non-zero elements correspond to true connections.
Value
truenet Reference (true) network of the example data set.
References
G. Altay, F. Emmert-Streib, "Inferring the conservative causal core of gene regulatory networks", BMC Systems Biology (2010) 4:132.
See Also
Examples
data(truenet)