HyphalNet

This project represents the implementation of community detection algorithms in biological networks. Our basic approach is shown below. It requires using an underlying protein-rotein interaciton network to derive sample-specific networks, that can then be compressed to a multigraph and reduced to functionally-specific communities.

We are currently evaluating this approach in three separate domains.

Install

To install the package, clone the repository.

1. git clone https://github.com/sgosline/hyphalnet.git
2. Then run: sudo python setup.py install

This will install the package and its files.

Getting Started

To run hyphalNet you need to download an interactome and create an igraph Graph and compress it using pickle. This script will do it for you.

1. Download the latest STRING DB interactome
2. Download the mapping file
3. Run:

   python ./bin/graphTools.py --graphFile [stringfile] --graphSource=string --nodeMapping=[mappingfile] --dest=./data

This will load up the interactomes in the correct format (for now we are working across 3 formats)

Build Pan-Cancer Signatures

We have built a script that downloads data from the [CPTAC]() repositories using the Python package cptac. To construct cancer signatures from the available data.

To build these scripts, run the script

python bin/buildCancerSigs.py --refName test

This will show the available options to build new cancer signatures.

The output of this will be a hyphalNetwork object in pkl format that can be used to map to novel datasets.

General Signature statistics

To collect information about the signatures, particularly ahead of paper publication, we run the following script to collect gene enrichment statistics, and distance metrics within the cancer signature data itself. Some of these are more useful than others.

Network Mutual Information

We wanted to show that the signatures we were getting were unique to the data collected. To do that we evaluated the newtork mutual information statistic. This probably does not need to be run more generally.

python bin/getSigNMI.py --hyph test_hypNet.pkl

Signature gene enrichment

We also wanted to capture tio biological activity of the underlying cancer signatures. This statistic is of interest for others wanting to do more analysis downstream.

python bin/hyphEnrich.py --hyph test_hypNet.pkl

Applying signature to new data

Lastly we need to apply the signatures to novel datasets. This script expects a csv of proteomics measurements in tidied format.

python bin/mapSigsToData.py --hyph test_hypNet.pkl