NetWalk and EF Tables

NetWalk analysis of selected data columns converts gene-centric values to interaction-centric values, which are derived by a random walk scoring of each interaction in the network for their relevance to the data based on the provided data values of genes as well as their network connectivity. Results of a NetWalk analysis are displayed in an Edge Flux Table (EF Table) in a TableView, and they are interpreted the same way as one would interpret gene expression data: positive Edge Flux values are associated with increased gene expression, and negative Edge Flux values are associated with decreased gene expression. For this section, we will use the normalized dataset we created at the end of the tutorial on Data import and processing.

  1. Select the Names column, and click on  to set this column as gene names.
  2. Click “Run NetWalk” button to display the Run NetWalk dialog.

    The list on the left shows data columns of the TableView. These can be selected to use as weights for NetWalk analysis, and a separate NetWalk will be run for each selected data column to generate EF Table. The data columns to be used as weights for NetWalk should be ratio-like values centered around 1, and should not contain negative values. Those data columns unsuitable for NetWalk analysis are colored red in the Run NetWalk dialog. Values to the right show cutoff percentages for data in the data columns to use as higher and lower cutoff of weights for NetWalk (see Manual for details). In most cases, these should not be modified. If “Generate FunWalk Table” is selected (default), a FunWalk will also be run in the end of NetWalk to generate FunWalk Table (see Manual for details). While EF Table has interactions in each row and EF values as their scores, FunTable contains functional terms (GO Terms) in each row of the table and their FunWalk scores in the rest of columns (see next section on FunWalk and FunTables).
  3. Select the 8 columns for Ratio values (those whose names start with “Ratio”). Click OK to run NetWalk analysis. Click OK again when prompted. Wait till NetWalk run finishes, it may take a while.
  4. After several seconds, two new TableView windows will be created, one containing EF Table and the other containing FunTable corresponding to the 8 conditions in the doxorubicin dataset.

     

    You can click on their corresponding nodes in the Object Tree to change their labels in the Properties panel or add notes about them.

    Any selected rows in an EF Table of a FunTable can be exported to a network view of genes involved in corresponding interactions or function terms, respectively. Just as an exercise, toggle row-selection, select a few interactions from the EFTable and click “Plot Network” button to generate a network in a NetView window.

    An EFTable can be processed just like a primary dataset (see above in Dataset import and processing). Next, we will show how to create EF Heatmaps for comparative analyses of networks between conditions. First, we will filter the EF Table to only show rows that have high EF values in either 1uM or 10uM condition at 24 hours.

  5. For this, we will use the “Filter rows” feature in TableView (note that this feature is present in all TableView windows, whether they contain primary datasets, EFTables or FunTables). Click on “Filter rows” button, this will bring up the dialog for designing row filters.

    The drop-down menu on top (next to “Filter by column:”) is to select the column according to the values of which rows will be sorted. Select the column for “Ratio: Doxorubicin 10uM; 24h/…”. You will see the drop-down box next to “Set Values” get activated, as this column is a numeric column. Select “>”, or larger than, and enter 1.5 in the text field next to it. What we did here is: we told the Row Filterer to only show rows in the current TableView where values in the “Ratio: Doxorubicin 10uM; 24h/…” column have values greater than 1.5. Do not change the “Clause to use with the current filter”. Press “Set as new Filter” button. Now, only rows where “Ratio: Doxorubicin 10uM; 24h/…” column is >1.5 are shown. We also want to include rows where EF values for the column “Ratio: Doxorubicin 1uM; 24h/…” are also high. So click on “Filter rows” again, and select ““Ratio: Doxorubicin 1uM; 24h/…”, enter 1.25 in the text field next to “Set Values:” (1uM 24h condition has less dynamic range than 10uM 24h condition, so we use a smaller cutoff of 1.25 to have equal number of rows for each). In the drop-down box next to “Clause to use with the current filter:”, select “OR” and hit “Add Filter”. This procedure creates a composite filter with the one we created earlier, so that now only rows that have >1.5 value in the “Ratio: Doxorubicin 10uM; 24h/…” column OR >1.25 value in the “Ratio: Doxorubicin 1uM; 24h/…” column are shown.

  6. Select the names column and the 8 original columns of EF Values, and click on  to make a new table for a clustering heatmap. Note that each time we want to make a clustering heatmap, we generate a new table of rows and columns that we wish to cluster. This is important, as the original table, though only showing the desired rows and columns, still contains the whole data table, therefore they are not suitable for clustering. Clustering analysis should only be done on an unfiltered dataset.
  7. Perform clustering and heatmap visualization as described in Clustering and Heatmap analysis. This is what you should see after resizing rows and columns (use “Resize”):

    Note that “Resize” works on all rows, but only on selected columns.

  8. Toggle row-selection, and select a number of rows from one of the clusters (e.g. the one where it shows increase in response to high dose but decrease in response to low dose) and click on “Plot Network” to plot the network corresponding to selected interactions.
  9. Select all nodes, go to “Coloring” in the NetView toolbar, select “Doxorubicin_MCF7.txt” as dataset, and “Log2 ratio: Doxorubicin 10uM; 24h/…” for column. Enter -1 on the left side of the drop down box with the color panels under “Color key”, and enter 1 on the right side. You can select any color panels in the drop down box. The coloring of selected nodes in the network will now reflect their relative expression values in the “Log2 ratio: Doxorubicin 10uM; 24h…” condition in the original dataset, according to the color key and data range you selected.
 

Now you can proceed to the next section on FunWalk and FunTables or to the section on Handling and analyzing networks.