heinz-cytoscape-app

This is a Cytoscape 3.x app (plugin) for discovery of subnetwork modules of differentially expressed genes in interaction networks. It provides an interface to the tool Heinz, which implements an algorithm that finds provably optimal subnetworks using a statistically interpretable gene-wise score. It runs the Heinz tool via a simple client-server protocol (https://github.com/melkebir/server-wrapper), so Heinz does not have to be installed on the computer from which the analysis is performed.

It is currently still in an early stage of development.

Authors

• Fedde Schaeffer
• Mohammed El-Kebir
• Gunnar W. Klau

Installation

The app is packaged using Maven. If Maven is installed, the app can be built from source by downloading (or cloning) the package from the code repository on GitHub, opening a terminal in the root directory of the package, and executing the command mvn install. Maven will then create a JAR file named heinz-cytoscape-app-version.jar in the target/ directory.

Once the app is built, it can be installed into Cytoscape simply by moving or copying the JAR file into the $HOME/CytoscapeConfiguration/3/apps/installed/ directory. It should then appear in the Apps menu in Cytoscape.

Running the servers

The app needs to contact a server to run Heinz. This server can be run locally, requiring Heinz (see https://software.cwi.nl/heinz) and an interpreter for the Python programming language (version 2.x, see https://www.python.org/) to be installed. The server wrapper can then be downloaded (or cloned) from https://github.com/melkebir/server-wrapper, and run using a command such as the following:

python src/server.py 9001 1 /path/to/heinz

Where 9000 is the port to accept connections on, 1 is the maximum number of Heinz runs to accept simultaneously, and /path/to/heinz is the path to the binary executable of Heinz.

To fit BUM models, the app requires another server to be running--possibly on the same machine, but a different port number. This server uses the same wrapper script, but instead of Heinz, it should run the R script fitBumModel.R provided in this package. The R script requires an the R environment for statistical computing to be installed, and the optparse package within R. Run R in a terminal and type install.packages("optparse") to install it, before quitting R with q(). Then run the following command to test if the script works:

Rscript /path/to/fitBumModel.R --help

On a *nix machine, the server can then be run with the command:

python src/server.py 9000 1 /path/to/fitBumModel.R