General Topological Overlap Measure

This package takes an undirected unweighted scipy.sparse adjacency matrix as an input and computes the GTOM(m) method, using m+1-step neighbors (1). It's highly efficient and can be used for parallel computation by calling the function for only few nodes at a time.

(1) Gene network interconnectedness and the generalized topological overlap measure, A. M. Yip and S. Horvath, BMC Bioinformatics 2007 8:22

Install

$ pip install gtom

Example

Recreating Panels a and c of Figure 3 of the technical report [1]:

#!python

import matplotlib.pyplot as pl
import networkx as nx
from gtom import gtom
import scipy.sparse as sprs
import numpy as np

edges = [(0,1),(1,2),(0,3),(0,4),(0,5),(0,7),
         (1,3),(1,4),(1,6),(1,8),(1,9),(1,10),
         (5,6),(7,8)]
G = nx.Graph()
G.add_edges_from(edges)

pos = nx.spring_layout(G)
labels = { n:str(n+1) for n in G.nodes()}
nx.draw(G,pos=pos)
nx.draw_networkx_labels(G,pos=pos,labels=labels)

N = G.number_of_nodes()
edges = np.array(edges,dtype=int)
A = sprs.csc_matrix((np.ones((edges.shape[0],)),(edges[:,0],edges[:,1])),dtype=float,shape=(N,N))
A += A.T

print("recreate results from figure 3 in [1]")
print("       |\t(i,j)=(1,2)\t(i,j)=(1,3)\t(i,j)=(2,3)")
print("---------------------------------------------------------")
for m in range(3):
    T = gtom(A,m)
    print(" m = %d |\t%f\t%f\t%f" %(m,T[0,1],T[0,2],T[1,2]))

pl.show()