Validation

PR- and ROC-Curves

PR- and ROC-curves are a classical way of measuring the performance of a network with respect to a known set of interactions. Those curves can be computed and plotted using the free open-source R platform and packages such as library(minet) or library(ROCR). The set of known interactions extracted from REDfly is available in the file redfly.data.

Coregulation statistics

In order to compute coregulation statistics, here are some C++ scripts able to deal with large networks (up to hundreds of thousand nodes).
The file fly.h and the files supervised.txt and unsupervised.txt, used in this page, are avalaible on the page Inference.
The following code is distributed freely in the hope that it will be useful but without any warranty (License GPL3)

  1. Count of coregulated genes with PPI.
    1. download the file ppi.data and place it in a data/ directory
    2. download the script tf-ppi-count.cpp
    3. compile g++ -o tfppicount tf-ppi-count.cpp
    4. execute ./tfppicount data/ppi.data supervised.txt results/tfppicount.txt in order to produce the file tfppicount.txt that compute the average number of PPI by tf.
  2. Average correlation of coregulated gene expressions.
    1. download the file rnaseq2.data and place it in a data/ directory
    2. download the script coreg-cor.cpp
    3. compile g++ -o coreg-cor coreg-cor.cpp
    4. execute ./coreg-cor data/rnaseq2.data supervised.txt results/coreg-cor.txt in order to produce the file coreg-cor.txt that lists the coregulated targets (with at least 50% of TFs in common) and their respective correlation of expression.
  3. Average Jaccard Index of a set of terms of coregulated genes.
    1. download the file imago.data or the file go.data and place it in a data/ directory
    2. download the script coreg-jc.cpp
    3. compile g++ -o coreg-jc coreg-jc.cpp
    4. execute ./coreg-jc data/imago.data supervised.txt results/coreg-jc.txt in order to produce the file coreg-jc.txt that lists the set of coregulated targets (with at least 50% of TFs in common) and their respective Jaccard index.
By dividing those numbers with those of random networks of similar structure, the respective enrichment value is directly computed.


If you have questions regarding the code, please send an email at software@meyerp.com.