Pathway Reconstruction

This repository contains the files needed to run and test various methods for signaling pathway reconstruction, with a focus on HybridLinker. There are thee things you need to do to get this code up and running:

1. Update the repo submodules
2. Put the interactome in the proper place
3. Install dependencies (including method-specific dependencies if you plan to run them)

Instructions for checking out this repo

There are three submodules in this repo. After cloning, you need to update the submodules with

git submodule init
git submodule update

Check the .gitmodules file for more details.

Adding the interactome

Within the localized-pathlinker directory, the interactome is in a zipped file. We need to unzip it and move it to a place where all methods expect to see it.

Locate and unzip the file.

cd Reconstruction/Methods/localized-pathlinker/Data
unzip PathLinker_2018_human-ppi-weighted-cap0_75.txt.zip

Make an Interactomes/ directory and move the interactome. Unfortunately a symbolic link won't work here.

mkdir ../../../../Interactomes/
mv PathLinker_2018_human-ppi-weighted-cap0_75.txt ../../../../Interactomes/

Installation Requirements

Most of the code in this repository is written in Python3. A full dependency list is not available at this time, but at least:

• NetworkX
• Pandas
• Numpy
• GraphSpace

Some of the code in this repository is written in Bash. Because of this, we cannot guarantee complete functionality on Windows systems at this time.

Install Python modules

The requirements are listed in requirements.txt. To install the python modules to your user directory (in python3), use the command

pip3 install --user -r requirements.txt

Depending on your install, you might use pip instead of pip3.

Alternatively, install dependencies in a new environment using conda with

conda env create -f depends.yml

Method-Specific Installation Requirements

Install CBC (ResponseNet)

CBC (Coin-or brank and cut) is an open-source linear programming solver written in C++. It is accessed from Python scripts using the mip python module. CBC is only needed if you plan to run ResponseNet, which relies on a linear program solver.

On a Mac OS X, the easiest way to install CBC is through Homebrew:

brew tap coin-or-tools/coinor
brew install cbc

For other platforms, see their website with instructions to download or build from source.

Install the Omics Integrator 2 (PCSF)

The most recent implementation of the Prize Collecting Steiner Forest (PCSF) is within the Forest module of the Omics Integrator v2. This is the maintained version of the code, which can be installed by cloning the repo and running setup.py. This is not included as a submodule for the reasons listed below; so make sure the OmicsIntegrator2 is included in your PYTHONPATH.

git clone git@github.com:fraenkel-lab/OmicsIntegrator2.git
cd OmicsIntegrator2/
python3 setup.py install --user

Depending on your distribution, this may not work the first time. We discovered that there is no pcst_fast module in Python3 (though it exists in Python2 via pypi). To remedy this, we

1. Installed pcst_fast from source.
2. Added pcst_fast to the PYTHONPATH environment variable.
3. Disable pcst_fast from the requirements in OmicsIntegrator.

These details are described below.

1. Installed pcst_fast from source, fixing the install so Python3 libraries are called.

git clone git@github.com:fraenkel-lab/pcst_fast.git

In the Makefile, I had to make the following changes to the pcst_fast_py line so python3 libraries were called:

• Added -undefined dynamic_lookup
• Changed python-config to python3-config

The line now looks like:

pcst_fast_py: $(PCST_FAST_PY_SRC_DEPS:%=$(SRCDIR)/%)
	$(CXX) $(CXXFLAGS) -undefined dynamic_lookup -shared -I $(SRCDIR) -I external/pybind11/include `python3-config --cflags --ldflags` $(SRCDIR)/pcst_fast_pybind.cc $(SRCDIR)/pcst_fast.cc -o pcst_fast.so

You can then install pcst_fast with

make pcst_fast_py

You should now be able to import pcst_fast in your Python environment.

2. Added pcst_fast to the PYTHONPATH environment variable. In your ~/.bashrc or ~/.bash_profile file (for linux and macs, respectively):

export PYTHONPATH=$PYTHONPATH:/Users/aritz/Documents/github/pcst_fast/

Remember to source ~/.bashrc or source ~/.bash_profile for the change to take effect.

3. Disable pcst_fast from the requirements in OmicsIntegrator (since we satisfied the requirement with #1 and #2). In the setup.py file, comment out the pcst_fast requirement. Lines 19-29 should now look like

install_requires=[
        "numpy",
        "pandas==0.23.4",
        "networkx==2.1",
        #"pcst_fast==1.0.7",
        "python-louvain",
        "goenrich",
        "sklearn",
        "axial",
        "scipy"
    ],

In the OmicsIntegrator/ directory, run

python3 setup.py install --user

(Alternatively, you could comment the pcst_fast requirement in requirements.txt and install those dependencies. We didn't check this).

You should now be able to import OmicsIntegrator in your Python environment.