Networks

Three networks are available for exploration:

1. The Core Network gives a comprehensive overview of the oligodendrocyte differentiation process. It was layouted and scientifically curated in a manual fashion and is well-suited for exploration.
2. The Differentiation Network was built automatically from the Core Network by adding a filtered set of reactions that are described in databases or derived from ChIP-Seq data. Filters included an expression threshold, association to a compiled list of Gene Ontology (GO) and Medical Heading Subjects (MeSH) terms, and association to identified active modules. We manually arranged the positions of the four major transcription factor hubs (Sox10, Tcf7l2, Olig2, and Myrf) and their shared targets to highlight patterns of joint regulation.
3. The Exploration Network was created in the same manner as the Differentiation Network but without the filtering procedure. Consequently, its factors and reactions are a superset of the Differentiation Network. The layout follows a gradient of factor connectivity: factors that participate in many different reactions are located in the center of the network, while factors with few reaction partners are situated towards the rim.

The controls for all three networks are identical and are described below. The expanded versions are appropriate for online and offline computer-aided analysis.

Pre-loaded data sets

You can select from 6 pre-loaded data sets up to 4 for simultaneous visualization. The pre-loaded data sets are based on publicly available data, and the sources for each category are accessible via links behind the category names.

The factors in the network view will morph into pie charts while at least one data set is selected. The expression of each factor is shown by projecting a color that is based either on logarithmized intensities (transcriptome) or logarithmized FPKM values (miRNome) into the pie slices. Each slice corresponds to one of the selected data sets - you can see which data set goes into which slice by checking the pie icons that appear to the left of the checkboxes. With each selected data set, the pie charts will acquire a slice until you exhaust the number of possible slices at four.

For details about the expression color scale, see the description below.

Adding your own data sets

You can add your own data to the network for data mining purposes. Due to the way this is implemented, no data upload to our server is necessary. Your data is only ever stored in your own browser session.

To successfully add data, you must first prepare a CSV file. This can be done in a spreadsheet application like LibreOffice or Microsoft Excel. The spreadsheet should look similar to this one:

• The table must have a header in the first row, and the header of the first column must be "Symbol". The headers of the remaining columns can be freely chosen , with each one representing one condition or difference.
  Column headers must be unique across all files you upload in one session.
• The first column lists the gene symbols for which you have data. As the map was set up to permit the analysis of human data, you need to employ the official HGNC symbols. (These are often identical to their mouse homologs' official symbols, but there are notable exceptions.) Gene symbols can be found in many other databases - see the below NCBI Gene and UniProt screenshots for eotaxin (CCL11) for reference:
  
  The page will automatically match up your symbols with all the corresponding factors (if there are any).
• All columns after the first list the values you want to project, with one condition or difference per column.
  There is no restriction on the kind of values you use, apart from them having to be numbers. The interpretation, of course, is yours to make.

Once you have the spreadsheet prepared, export a file in CSV format. Then click on the very first list entry Your data (click to add) and select the exported file. Your column names will be appended as data sets to that entry and can be selected for visualization from there.

Highlighting factors via ontology associations

We have associated each factor in the network with the GO term and Reactome ontologies. Ontology terms that were related to at least three genes in the current map can be selected from the drop-down menu, and the associated factors will be highlighted with a golden halo.

Legend colors and how to adjust them

While no data set is selected, the factors in the current map are colored according to their molecular type. Some will also have individual shapes. In this state, the legend shows a color code for reaction and factor types.

On selecting a data set, the legend will adapt to represent the spectrum of expression values that were observed in the selected categories.

The expression color legend comes with two sliders that can be used to adjust the colors in the network. The sliders initially occupy their respecive maxima, i.e., lowest and highest observed expression. Moving either slider closer to zero will increase the saturation in all like-colored pie slices, while moving them towards the limits will decrease it. Note that all factors with absolute expression higher than the adjusted slider position will drop out of the linear color projection range and appear fully saturated.

Exploring subnetworks from our case study

We applied Cytoscape and its plug-in jActiveModules and KeyPathwayMiner to extract activated subnetworks from the Exploration Network. These represent network regions that are enriched in expressed factors. For details of the procedure, please refer to the corresponding user manuals.

The fourth option removes all ChIP-derived interactions to give an impression of the remaining network.

Select an entry from the drop-down menu to view the corresponding activated subnetwork. To return to the original network, remove the selected entry from the drop-down menu by clicking on the × icon.

Note that activated subnetworks are only available for the Exploration Network.

Extracting customized subnetworks

You can focus on your molecules of interest by eliminating all other factors from view. To do this, first mark all relevant factors via either of two methods:

• Click on a factor in the network view and press the Mark this factor button.
• Select the factor from the second drop-down menu below the one for regulatory subnetwork.

No matter which method you use, every marked factor will be added to the above-mentioned drop-down and shown with a thick black rim in the network view. Factors can be unmarked by deleting them from the drop-down or by clicking the Unmark this factor after seleting it in the network view.

Once you are finished with your selection, click the Extract button at the bottom right side. Its label will change to Restore and all unmarked factors will be temporarily removed from the network view. If the ☑ Include connected factors option above the button is checked, factors with direct connections to the marked ones will remain visible.