Locked History Actions

Diff for "PropheticGrangerOptions"

Differences between revisions 1 and 2
Revision 1 as of 2015-02-19 11:22:21
Size: 3277
Editor: OriolGuitart
Revision 2 as of 2015-02-19 11:33:53
Size: 3323
Editor: OriolGuitart
Deletions are marked like this. Additions are marked like this.
Line 1: Line 1:
== PropheticGranger Selection Options ==

PropheticGranger Selection Options

As stated in PropheticGranger algorithm presentation, this method contains two independent stages: computation of a biological prior and inference through Granger-causality-based regression. PropheticGranger allos users to apply only one of this stages or both. This can be selected through the Algorithm Mode parameter. Depending on the chosen option for this parameter, the rest of input parameters will be available for selection or not. Figure below shows the available mode options.

TDC Algorithm modes

If the chosen mode includes the use of network prior information, then users have two options regarding this prior information.

  1. Use Human Pathway Commons database prior data: PropheticGranger provides direct acces to Human Pathway Commons database so if your data is human related, you only need to provide the column in your data that contain the HUGO ids that will allow to map your data with the known human pathways.

  2. Use your own set of network prior data: Users might bring their own network prior data by just loading a zip file with all network files in sif format. At the same time, it is required to provide the column in the input data that will be used to map the data with the network prior data.

These two options are availabe through the Prior Data Mode parameter. Figure below shows these two options.

TDC Algorithm prior modes

Finally, if the chosen algorithm mode includes the use of the Granger-based regression method, the input data for this method needs to defined. This data is defined in Sources for Network Inference and is related to the Data Table selected on the top of the Cyni UI.

Granger-based regression method requires the input data to be formatted in a specific way. As any other regression method requires samples and features. The samples would be the Inhibitor dimension and the features would be the Time dimension. Therefore, the input data for this method needs to be formatted so that the data associated to these two dimensions can be easily extracted by the algorithm. Cytoscape only accepts two dimensional data so the way to input this kind of data is by using the column name to define the different dimensions. Moreover, there is the option to infer a different network from a possible third dimension. The figure below shows these two possible options in the input data UI. The selection panel only shows the data columns that have their name formatted in the predefined way selected for the parameter Data Columns Name Format.

Therefore, your data should be modified to follow these rules. For instance, if you only have a simple matrix such as Proteins by Observations, you should modify the column names so that new column names have the format: ObservationNames/0min. So it would be the same column names with a "/0min" suffix added. However, if you have other kind of more complex data, you will be able to take advantage of this notation to use all dimensions of your data for the inference process.

TDC Algorithm input data selection