TieDIE - Tied Diffusion of Interacting Events Algorithm
The TieDIE algorithm computes a subnetwork of gene and protein interactions that connect genomic perturbations to gene expression changes characteristic of cancer subtypes. The pathways derived in this way predict interlinking genes that may correspond to essential components of cancer signaling and may provide a mechanistic explanation of tumor behavior and suggest subtype-specific drug targets. In addition, the algorithm can extend to generate connecting subnetworks for generic gene sets, and runs a null-model permutation test to determine if these sets are significantly close in pathway space.
Source code is available from the public git repository:
Also, a full tutorial, using a simple test example, is attached here, as a PDF.
Comments, and feature requests may be sent to <firstname.lastname@example.org>.
TieDIE Input Format:
pathway.sif: A pathway in .sif format. This format iscompatible with Cytoscape, and uses a 3-column tab-separated <source> interaction> <target>. For example:
MDM2 inhibits> TP53
TP53 activates> MDM2
upstream.input: Input heats for the `upstream" set of genes, weighted by frequency of mutation or amplification. The 3-column tab-separated format
should contain a column for the gene name, the input heat, and the expected functional effect of the perturbation (+/-). For example:
PIK3CA 0.5 +
RB1 0.3 -
TP53 0.2 -
downstream.input: Input heats for the `downstream" set of transcriptional responses. The format is the same as the upstream.input file, but the third column should be interpreted as the inferred activity.
A test example is available to help run TieDIE; see the README.md file for the relative location in the github directory. To run the test, change directory to this folder and type `make". The program should run in a few seconds and report to stderr, a sub directory with the output will be created. Important output files are:
Visualizting TieDIE Solutions:
There are many software packages available for network visualization, but we've found the Cytoscape package to be well suited for network visualization. The TieDIE output network (tiedie.sif) can be directly imported to Cytoscape 2.8 or later, and the node attribute file (heats.NA) along with the supplied properties file (vizmap.props) can be used to color nodes by linker heat, as shown in the figure below.
More detailed visualizations can also be performed in Cytoscape: for example, nodes can be colored or shaped based on source or target node status, and nodes/edges can be highlighted. Additional examples can be found in the full tutorial, which is attached in this page as a PDF.
The more complex images shown in the TieDIE paper (Figures 4,5) allow visualization of node-specific data within an embedded network view, and are referred to as 'circleplots'. This code is available at https://github.com/ ucscCancer/paradigm-scripts and requires the original perturbation-matrix and gene-expression data to run. The circlePlot.py program produces a set of PNG format images that can be imported to Cytoscape. Unfortunately, there's no simple way to bypass the GUI loader of Cytoscape and programatically load images, although this is rarely an issue with small networks.
Much more detailed information is available in the full tutorial, which is attached as a PDF document to this page. In addition, the README file on github provides information on the directory structure of the code, and unit and regression tests are provided for those who wish to modify the code.