Create and visualize a module network of regulatory genes
Added by GenomeSpaceTeam on 2015.05.22
Last updated on over 3 years ago.
Does my gene expression dataset contain a module network of regulatory genes? Does the network have any special features?
This recipe provides one method for creating and visualizing a module network of regulatory genes. An example use of this recipe is a case where an investigator may want to evaulate an expression dataset to find regulatory genes such as transcription factors, and then determine if they are connected in a network.
In particular, the regulatory genes of interest are genes which regulate other genes associated with an embryonic stem cell (ESC) state. This 'stemness signature' is a feature common to ESCs, as well as induced pluripotent stem cells (iPSCs), and also in a compendium of human cancers, such as breast cancer. This recipe recapitulates research by Wong et al., in Cell Stem Cell (2008), "Module map of stem cell genes guides creation of epithelial cancer stem cells."
We use a gene expression dataset of primary human breast cancer tumor samples (described in Chin, K. et al, Cancer Cell, 2006), and create a module network by projecting a set of stemness regulators onto the gene expression dataset, using Genomica. A module network is a model which identifies regulatory modules from gene expression data, especially modules of co-regulated genes and their regulators. The module also identifies the conditions under which the regulation can occur.
After obtaining the module network, we visualize it using Cytoscape. Since the network is very large, we then filter it to just a subnetwork of stemness regulators and their connections, again using Cytoscape. This provides us with a visual representation of the stemness regulators as they appear projected onto a breast cancer tumor dataset.
To complete this recipe, we will need a gene expression dataset and a list of genes which we believe to be important transcriptional regulators. In this example, we use a gene expression dataset of primary human breast cancer tumor samples, which is fully described in Chin, K. et al, Cancer Cell (2006). We also use a set of genes which are believed to be regulators of the embryonic stem cell state, called "stemness regulators". This gene set of stemness regulators can be obtained by following the recipe for completing Stepwise Linkage Analysis of Microarray Signatures (SLAMS). We will need the following datasets, which can be downloaded from the following folders:
breasttumor.preprocessed.collapsed.tab: this file contains the gene expression profile of primary human breast cancer tumor samples. The original dataset has been log-transformed, row-centered on the mean, and has had the probe IDs collapsed to HUGO Gene Symbols.
stemness_regulators.geneset.tab: this file contains a list of genes which are believed to regulate genes associated with the embryonic stem cell state.
A subnetwork of genes which regulate the transcription of embryonic stem cell-associated signature genes. This subnetwork shows the connections between regulators of this 'stemness signature'.
First, we will build a network of regulatory modules from co-regulated genes with similar expression profiles, from the breast cancer tumor sample dataset.
breasttumor.preprocessed.collapsed.tab) by clicking and dragging the file to the Genomica tool icon.
Algorithms > Create a Module Network…
Maximum tree depthto 5.
Candidate regulator genes, click
stemness_regulators.geneset.tab), then click
Runto create a module network.
GenomeSpace > Export Network to GenomeSpace…
OKto close the pop-up and continue with the export (see below).
Next, we will import the network results from GenomeSpace into Cytoscape, and show the gene regulatory network. In this example, we use Cytoscape 3.4.
NOTE: This recipe requires Cytoscape 3.3 or higher to be installed. Please navigate to www.cytoscape.org to download the latest version.
cytoscape.jnlpfile. Double-click this file to launch Cytoscape.
Start New Sessionlabel, choose
With Empty Network.
NDB Readerapp. To install this, use the following steps:
Apps > App Manager.
Installto install it. Once the app is installed it allows you to load NDB files, without modifying any of the Cytoscape menu items.
File > Import > Network > GenomeSpace.
Selectto load the file. This may take a few moments. Once the file is loaded, it will prompt you to import the network to a network view. Choose a network and click
OKto create the network. This may take a few minutes.
This will create a network of 13,341 nodes and 117,815 edges. This is an incredibly large, dense network. It is best to filter the network before attempting to visualize it using a different Cytoscape layout, such as a force-directed layout or a degree-sorted circular layout. In the next step, we will filter the network to a more manageable size.
In this step, we will filter the existing network of 13,341 nodes to just the nodes that are known to be regulators of the embryonic stemness signature. We will filter by the gene set described in
File > Import > Table > GenomeSpace.
Select. This will bring up an import menu, allowing you to choose which names and symbols are imported into Cytoscape.
OKto import the file.
Control Panelsection, click the
Default Filtersection, click the
+button to add a new filter, and choose
node.INGENESETas the attribute to filter on.
New Network From Selection, All Edges. This will create a new subnetwork with just the stemness regulators and their connections. You should obtain a network of 48 nodes and 334 edges.
Layout > Edge Weighted Spring Embedded Layout
Layout > Circular Layout
Layout > Degree Sorted Circle Layout
Layout > Scale, then use the sliding scale bar to increase or decrease network density.
Styletab in the
Control Panel. You can choose preset styles using a drop-down menu. You can create your own styles by clicking on the
Defaultspane, then adjusting the parameters.
Edgetab in the bottom part of the pane. Then click on
EDGE_COLOR, choose a new color, and click
OK. Some variables take numeric inputs, such as
This is an example interpretation of the results from this recipe. First, we created a module network of genes that regulate the stemness signature, by projecting it onto a breast cancer tumor gene expression dataset using Genomica. We then used Cytoscape to visualize the connections between these genes, and to filter a large network down into a subnetwork of just stemness regulators.
We can see from the resulting subnetwork (see below) that, e.g. genes such as POLR2K and MYC are stemness regulators with many connections to other stemness regulators. In contrast, genes such as SQLE or PSCA have fewer connections to other stemness regulators. We can describe POLR2K and MYC as having higher 'degree', than genes such as SQLE or PSCA. These connections in the network imply that perturbing the regulation of POLR2K or MYC would have a larger effect on the embryonic stem cell expression signature (because their multiple connections allow perturbations to propagate in the network) than perturbing other genes that have fewer connections.