Modulated Sub-graph Finder (MSF) used to find the significantly dis-regulated sub-graphs or cluster of genes from the host cell signaling network, giving these sub-graphs an overall significance of modulation by combining the individual p-values of the genes derived from differential genes expression analysis.
- Java version 8
- Jdk 1.8
Download the ModulatedSubgraphFinder.jar file and the example files for tutorial.
To find the Modulated sub-graphs you need two files. One file is tab-separated text file containing the output of DESeq2/EdgeR analysis. Second file is a directed adjacency list with 3 columns, first two columns with interacting genes and the third column direction of the interaction. The gene identifiers should be same in both the input files. Example files included in Docs folder.
MSF has seven argument parameters
-pThe path to differential gene expression analysis file-iThe path to network file (Interaction file)-tSoftware used for differential gene expression analysis (DEseq2 or EdgeR)-eThe extension limit (1 to 3 genes extension)-mThe merging limit (1 to 3 genes merging)-kOutput extra files (InitialGraphs, ExtendedGraphs and MergedGraphs)-oThe path to output folder
Navigate to folder containing the jar file and the example files and run command
java -jar ModulatedSubgraphFinder.jar -p ExampleDEGAnalysis_DEseq2.csv -i ExampleInteractions.csv -t DEseq2 -k yes -o /home/Documents/
The default extension and merging limit is 2. If you do not want to output extra files skip the -k argument.
This is a text file that contains the initial sub-graphs that are found by combining the individual P-values of the gene. The snapshot shows there core sub-graphs were found, written in a line with the combined p-value of the sub-graph at the end.
[ppp2r1a, tgfbr2, tgfb2, ppp2ca] 1.1103580381200124E-18
[smad2, skp1a, smad3] 1.6569054494446366E-7
[nog, bmp2, bmp5, bmpr2, gdf7, bmp6] 3.482452728268697E-4
This file shows if any sub-graphs were extended by adding genes beyond its immediate neighborhood. For example graph is extended by 4 genes.
[ppp2r1a, tgfbr2, tgfb2, ppp2ca] 1.1103580381200124E-18
[smad2, skp1a, smad3, smad7, ifng, acvr1, smad5] 1.9580322709452766E-10
[nog, bmp2, bmp5, bmpr2, gdf7, bmp6] 3.482452728268697E-4
This is the text file showing if any or all the (extended/unextended) sub-graphs merge with each other or not. In this case the graphs did not merge.
[ppp2r1a, tgfbr2, tgfb2, ppp2ca] 1.1103580381200124E-18
[smad2, skp1a, smad3, smad7, ifng, acvr1, smad5] 1.9580322709452766E-10
[nog, bmp2, bmp5, bmpr2, gdf7, bmp6] 3.482452728268697E-4
This output file gives details about the genes found in the sub-graphs. It shows graph number followed by genes in the graph. Then each gene from the graph, its fold change, individual p-value and in the last if it was identified as a source, intermediate or sink. The sources have impact score against it.
[Graph 1]
[ppp2r1a, tgfbr2, tgfb2, ppp2ca]
[ppp2r1a, 343.680151982419, 1.4E-7, Source, 3/4 = 75.000]
[tgfbr2, 154.225593438929, 0.03294361100258, Intermediate]
[tgfb2, 505.093869916889, 4.0E-5, Sink]
[ppp2ca, 525.820339004501, 6.44660834994E-4, Source, 3/4 = 75.000]
[Graph 2]
[smad2, skp1a, smad3, smad7, ifng, acvr1, smad5]
[smad2, 28.1552749566574, 1.51528460481E-4, Intermediate]
[skp1a, 843.231746773892, 0.014525956708056, Sink]
[smad3, 176.442049495414, 0.146036813182214, Source, 2/7 = 28.571]
[smad7, 606.619477359416, 0.562318600418486, Source, 4/7 = 57.143]
[ifng, 135.622073938663, 2.12566137193E-4, Sink]
[acvr1, 100.463207564239, 0.017364084706018, Source, 2/7 = 28.571]
[smad5, 675.586240911088, 0.005281642821289, Sink]
[Graph 3]
[nog, bmp2, bmp5, bmpr2, gdf7, bmp6]
[nog, 48.0012337399973, 0.005498079382534, Intermediate]
[bmp2, 186.03559350895, 0.237098851922717, Sink]
[bmp5, 57.9279489451268, 0.419057181261883, Sink]
[bmpr2, 814.343493775222, 0.017218841862483, Source, 2/6 = 33.333]
[gdf7, 142.559614133974, 0.011845850200211, Source, 5/6 = 83.333]
[bmp6, 442.598787102063, 0.403941986198972, Sink]
A text file with directed adjacency list for MSF identified modulated sub-graphs. This file could further be used to visualize the sub-graphs in other tools for example in Cytoscape. The last two columns are used as edge attributes to be imported in Cytoscape.
tgfbr2 ppp2r1a |- 0 1
tgfb2 tgfbr2 <- 0 1
tgfbr2 ppp2ca |- 0 1
skp1a smad2 <- 0 1
skp1a smad3 <- 0 1
nog bmp2 -> 1 0
nog bmp5 -> 1 0
bmp2 bmpr2 <- 0 1
nog gdf7 <- 0 1
nog bmp6 -> 1 0
smad7 smad2 -> 1 0
ifng smad7 <- 0 1
acvr1 smad3 <-> 1 1
acvr1 smad5 -> 1 0
This file has node attributes for each gene to be imported into Cytoscape. First column with gene name, second column the node size depending on the source weight and the last column is LogFoldchange to show type of regulation.
ppp2r1a 750.000 -1.15250657211695
tgfbr2 10.000 0.357347052992014
tgfb2 10.000 -0.273710254909285
ppp2ca 750.000 -0.225269184801659
smad2 10.000 1.20822323638527
skp1a 10.000 0.47396118683819
smad3 285.714 -0.193102843027487
smad7 571.429 0.100940948274761
ifng 10.000 -0.411195572576727
acvr1 285.714 -0.360351622623896
smad5 10.000 0.43197314616271
nog 10.000 -0.397877024648883
bmp2 10.000 -0.135915332971465
bmp5 10.000 -0.21104227566159
bmpr2 333.333 0.255224975603306
gdf7 833.333 0.334306828205764
bmp6 10.000 0.079033613236002
- Cytoscpae
- StringApp (Cytoscape Plugin)
- enhanceGraphics (Cytoscape Plugin)
File-> Import-> Network-> File-> NetworkFile.text-> Advanced Options-> Tick Delimiter **SPACE**-> Untick Use first line as column names
Click column 1 and select as source node Click on column 2 and set as target node Click on column 3 and set as interaction type. Click ok.
The MSF network has been imported to Cytoscape, to add directional follow the next steps
Click on Style Select Edge attributes Drop down Source Arrow Shape
Column -> Column 6
Mapping Type -> Discrete Mapping
For 1 select the arrow ->
Drop down Target Arrow Shape
Column -> Column 4
Mapping Type -> Discrete Mapping
For 1 select the arrow ->
To add node attributes follow steps
File-> Import-> Table-> File-> SourceWeight.text Where to import Table Data -> To selected netowkrs only Network List -> NetworkFile.text Import Data as -> Node Table Columns Key Column for Networks -> Shared name Advanced Options-> Tick Delimiter **SPACE**-> Untick Use first line as column names
Click on Style Select Node attributes Drop down size
Column -> Column 3
Mapping Type -> Passthrough Mapping
Drop down Border Paint
Column -> Column 5
Mapping Type -> Continous Mapping
Set Border width to 25
After successful download of Cytoscape and StringApp and enhanceGraphics inside cytoscape. Follow the steps
File->Import->Network->Public Databases
A new window would open then select Data source as String : protein query. Next select the organism as Homo sapiens. Then enter the complete list of gene in MSF sub-graphs and click import. The string network from String database would be downloaded with the user input gene list but interactions from String database. Since we would have our own interaction we can remove the String interactions by
select->Edges->Select all edges The right click on any edge in the network and click Edit->cut. All the string database edges would be removed.
We would overlay our network on the String network with our edges only. To do so
Tools->Merge->Networks a new window would open, using the Union option select the string network and MSF identified network. From the advance options select the appropriate names of genes to be merged For example SharedName for MSF network and display name for String network. Then click merge.
Now you would see MSF identified network overlapped with string network.
To perform gene enrichment analysis on MSF network got to Apps->STRING->Set as STRING network. Next go to Apps->STRING Enrichment->Retrieve Functional Enrichment. This would perform functional enrichment for GO terms and KEGG pathways. Below the network a table with enrichment analysis would open. You can filter the table to a particular type of enrichment for example KEGG pathway using the filter symbol on top left of table. Using the settings option on top right you could make single genes to pie charts to show for example in which KEGG pathways is the gene present with different colors for different KEGG pathways.
You can either save the network as a png or a session file.
The interaction file used to build and test MSF was downloaded from reactome https://reactome.org/download-data under Functional interactions (FIs) version 2016. This file was pre-filtered before use. All direct activation and inhibition interactions were filtered to be used by MSF testing.
First interactions that had no directions were removed from the file. Then indirect interactions were removed followed by interactions like expression regulation and catalyze. The remaining interaction were checked manually.
Eclipse Neon
Version 1.0 of the tool.
Mariam Farman farman@tbi.univie.ac.at
This project is licensed under the Creative Commons Attribution 4.0 International License.