Example uses of the Multi-Dendrix pipeline

I tested the runtimes of the examples below using a machine running 64-bit debian linux, with 24GB of available RAM, and a Xeon 2.4Ghz CPU.

Batch Multi-Dendrix

batch_multi_dendrix.run()

This is a simple Python script for running Multi-Dendrix on different datasets. In this example, I run Multi-Dendrix for k=3, t=3 on each of the three mutation datasets: GBM(2008), GBM, and BRCA.

On my machine, it takes ~20.15 seconds to complete.

Analyzing subtype-specific mutations

subtype_specific_mutations.run()

This is a simple Python script for computing p-values of subtype-specific mutations. The script requires mutation data and a list of patients and their respective subtypes, The default parameters here use the BRCA mutation data in the examples/mutation_data/BRCA directory.

On my machine, it takes 0.5 seconds to compute the subtype-specific mutations on the BRCA dataset.

Generating permutation data

generating_permuted_matrices.run()

This is a simple Python script for generating permuted mutation data. The default parameter values are for permuting the GBM(2008) mutation data in the examples/mutation_data/GBM_2008 directory, which is a small dataset. The main utility of this script is that it is much faster to generate permutation data before running the permutation test.

On my machine, it takes ~3 seconds to permute the mutation data once.

Weight permutation test

weight_permutation_test.run()

This script performs the matrix permutation test on a collection of gene sets. It gives an example of how one could use various functions from different modules in order to evaluate the collection without obtaining the collection from the Multi-Dendrix pipeline. The program first loads the GBM(2008) mutation data, permutes it the given number of times, and then performs the matrix permutation test.

The default parameters permute the GBM(2008) data ten times. On my machine, this script runs in 41.2 seconds.

Generating permuted networks

generating_permuted_networks.run()

This is a simple Python script for generating permuted PPI networks. The default parameter values are for permuting iRefIndex PPI network in examples/networks/ directory, which is a large PPI network. The main utility of this script is that it is much faster to generate permutation data before running the permutation test.

On my machine, it takes 14 minutes and 5 seconds to permute iRef once using Q=100, since this requires performing 21,274,600 edge swaps.

Direct interactions test

direct_interactions_test.run()

This script performs the direct interaction test on a collection of gene sets. It gives an example of how one could use various functions from different modules in order to evaluate the collection without obtaining the collection from the Multi-Dendrix pipeline. The program first loads the iRef network, permutes it the given number of times, and then performs the direct interactions test of the pathway set.

The default parameters permute the iRefIndex once, and only swaps | E | total edges (i.e. Q=1). On my machine, this script runs in 19.2 seconds.

Results from users

• Max Leiserson at Brown University