Proximity NetMed

Proximity is a package for the calculation of network-based distances like proximity and separation.

Proximity

First introduced in the paper "Network-based in silico drug efficacy screening" by Guney et al. 2016, proximity is a distance measure based on shortest paths between two sets of nodes in the protein-protein interaction network.

Separation

The separation between two set of nodes is a network-based measure introduced in the paper "Uncovering disease-disease relationships through the incomplete interactome" by Menche et al. 2015.

Implemented by: Rodrigo Dorantes Gilardi rodgdor@gmail.com for the Barabasi Lab.

Usage

In order to use proximity.py you need to clone the repository and add the module to your path as follows:

git clone https://github.com/Barabasi-Lab/proximity ~/proximity
EXPORT PATH="$HOME/proximity:$PATH"

Example

Here is an example to compute the proximity between a set of genes T and a set of disease genes S. The script assumes that the protein-protein interaction network is at '../test_proximity/data/ppi.csv' and that you have networkx, graph-tool, numpy, and pandas installed.

You can select the relevant network science module (networkx or graph-tool) and remove the code for the other module.

import proximity
import numpy as np
import pandas as pd
import networkx as nx
import graph_tool.all as gt

# Load the PPI
df = pd.read_csv('../test_proximity/data/ppi.csv').dropna()
df = df[['Symbol_A', 'Symbol_B']]

# Select some Targets and Disease genes
np.random.seed(42)
T = np.random.choice(df['Symbol_A'].values, 10, replace=False)
S = np.random.choice(df['Symbol_B'].values, 10, replace=False)

# Create the networks

# Graph-tool
g = gt.Graph(directed=False)
ids = g.add_edge_list(df.values, hashed=True)
g.vertex_properties['ids'] = ids

# Networkx
G = nx.from_pandas_edgelist(df, source='Symbol_A', target='Symbol_B')

# Compute proximity
net_nx = proximity.Network(g)
net_gt = proximity.Network(G)

print(net_nx.get_proximity(T, S, n_iter=1000))
print(net_gt.get_proximity(T, S, n_iter=1000))

This will output something like this (result may vary):

{'d_c': 1.8, 'z_score': -0.19576741426849495, 'mu': 1.8270333333333333, 'sigma': 0.13808903506411457}
{'d_c': 1.8, 'z_score': -0.5279439376809244, 'mu': 1.8639444444444444, 'sigma': 0.12111976268792904}