Differential network tests

When using the DNT package, please cite the accompanying paper, where also additional methodological information can be found:

Schefzik, R., Boland, L., Hahn, B., Kirschning, T., Lindner, H. A., Thiel, M., and Schneider-Lindner, V. (2022). Differential network testing reveals diverging dynamics of organ system interactions for survivors and non-survivors in intensive care medicine. Frontiers in Physiology, 12:801622. doi: 10.3389/fphys.2021.801622

Please make sure that you have the latest version of R installed.

To install the DNT package, run the following:
install.packages("remotes")

library(remotes)

remotes::install_github("RomanSchefzik/DNT")

The DNT package can then be loaded using
library(DNT)

The DNT package provides an overall frame for testing for differences between two statistical networks based on permutation tests. There are various options for (i) network estimation (i.e. specifying the adjacency matrices) and (ii) network comparison (using various types of overall, node-specifc or edge-specific network difference chacteristics). Moreover, tools for graphical illustrations and comparisons are offered. DNT fuses scattered functions from packages such as igraph, Hmisc, energy and qgraph in an overarching frame for differential network testing.

The DNT package essentially contains three main functions: create.adjacency.matrix, perm.test.nw and comp.plot.

With the function create.adjacency.matrix(A,methodlist,thresh), an adjacency matrix representing a network for an input data table A can be created, where the specific method for estimating the adjacency matrix can be specified in the methodlist argument. Currently, the following options are implemented: Spearman correlations, partial Spearman correlations, distance correlations (for all these options, edge weights are set to zero if the corresponding (adjusted or unadjusted) p-value is greater than the specified significance threshold thresh, which is typically set to 0.05) and EBICglasso.
See ?create.adjacency.matrix for details, in particular on how to exactly set up the methodlist argument.

create.adjacency.matrix(A=ExDataA,methodlist=list("Spearman"),thresh=0.05)
create.adjacency.matrix(A=ExDataA,methodlist=list("PCSpearman.adj","BH"),thresh=0.1)
create.adjacency.matrix(A=ExDataB,methodlist=list("EBICglasso","spearman",0.1))

The main function of the DNT package is perm.test.nw(A,B,permnum,methodlist,thresh,score.funct,paired), which allows to test for differences between two statistical networks derived from input data sets A and B in a testing frame based on a specified number of permutations permnum. Both a network estimation method and a network difference charcteristic, according to which the network difference is measured, has to be chosen. For network estimation and creation of adjacency matrices, the above-mentioned methods can be specified in the argument methodlist (and, accompanying, thresh). The considered network difference characteristic has to be specified in the score.funct argument, where several options for overall (Frobenius metric, difference in global strength, maximum metric, spectral distance, difference in the number of edges/clusters/isolated nodes), node-specific (difference in betweenness/closeness/degree/eigenvector centrality) or edge-specific (difference in (absolute) edge weights) network difference chracteristics are provided. Finally, in the paired argument, it can be chosen whether a paired or unpaired version of the permutation test frame shall be considered, depending on the structure of the input data. The output of the perm.test.nw function comprises the estimated adjacency matrices, the value(s) of the test statistic(s), the values of the test statistic(s) when applying permutations, the corresponding p-values and the corresponding p-values when including a pseudocount to avoid p-values that are exactly zero in permutation-based settings.
See ?perm.test.nw for details, in particular on how to exactly set up the methodlist and score.funct arguments, and for the precise output format of the function.

res1<-perm.test.nw(A=ExDataA,B=ExDataB,permnum=10000,methodlist=list("Spearman"),thresh=0.05,score.func=global.str,paired=TRUE)
res2<-perm.test.nw(A=ExDataA,B=ExDataB,permnum=10000,methodlist=list("EBICglasso","spearman",0.1), score.funct=degree.inv)
res3<-perm.test.nw(A=ExDataA,B=ExDataB,permnum=10000,methodlist=list("Spearman.adj","bonferroni"),score.funct=edge.inv)

The function comp.plot provides a tool for visual comparison between two networks, where the above-mentioned methods can be used for network estimation and creating the adjacency matrices. Moreover, several options for setting up the layout and the graphical representation are available.
See ?comp.plot for details.

comp.plot(A=ExDataA,B=ExDataB,methodlist=list("Spearman"))

comp.plot(A=ExDataA,B=ExDataB,methodlist=list("PCSpearman.adj","bonferroni"),layout=igraph::layout.circle,curved=FALSE)

comp.plot(A=ExDataA,B=ExDataB,methodlist=list("EBICglasso","spearman",0.1),layout=igraph::layout.fruchterman.reingold,curved=FALSE)

By calling create.app() in the R console, an R Shiny app in which the function graph.plot for visual comparison of two networks is implemented in a user-friendly interface will be opened.

create.app()