def calculate_lcc_significance(network,
nodes,
nodes_random=None,
bins=None,
n_random=1000,
min_bin_size=100,
seed=452456):
# Degree matching problematic for small bin sizes
#if bins is None and nodes_random is None:
# bins = network_utilities.get_degree_binning(network, min_bin_size)
if nodes_random is None:
network_nodes = list(network.nodes())
#nodes_random = get_random_nodes(nodes, network, bins = bins, n_random = n_random, min_bin_size = min_bin_size, seed = seed)
nodes_random = []
for i in xrange(n_random):
shuffle(network_nodes)
nodes_random.append(network_nodes[:len(nodes)])
network_sub = network.subgraph(nodes)
component_nodes = network_utilities.get_connected_components(
network_sub, False)[0]
d = len(component_nodes)
values = numpy.empty(len(nodes_random))
for i, nodes in enumerate(nodes_random):
network_sub = network.subgraph(nodes)
component_nodes = network_utilities.get_connected_components(
network_sub, False)[0]
values[i] = len(component_nodes)
m, s = numpy.mean(values), numpy.std(values)
if s == 0:
z = 0.0
else:
z = (d - m) / s
return d, z, (m, s)
def calculate_lcc_significance(network, nodes, nodes_random=None, bins=None, n_random=1000, min_bin_size=100, seed=452456):
# Degree matching problematic for small bin sizes
#if bins is None and nodes_random is None:
# bins = network_utilities.get_degree_binning(network, min_bin_size)
random.seed(seed)
if nodes_random is None:
network_nodes = list(network.nodes())
#nodes_random = get_random_nodes(nodes, network, bins = bins, n_random = n_random, min_bin_size = min_bin_size, seed = seed)
nodes_random = []
for i in xrange(n_random):
random.shuffle(network_nodes)
nodes_random.append(network_nodes[:len(nodes)])
network_sub = network.subgraph(nodes)
component_nodes = network_utilities.get_connected_components(network_sub, False)
#print component_nodes
d = len(component_nodes[0])
values = numpy.empty(len(nodes_random))
for i, nodes in enumerate(nodes_random):
network_sub = network.subgraph(nodes)
component_nodes = network_utilities.get_connected_components(network_sub, False)[0]
values[i] = len(component_nodes)
m, s = numpy.mean(values), numpy.std(values)
if s == 0:
z = 0.0
else:
z = (d - m) / s
return d, z, (m, s), values
def get_diamond_genes(network_file, seeds, file_name, only_lcc=True):
network = get_network(network_file, only_lcc=only_lcc)
nodes = set(network.nodes())
seeds = set(seeds) & nodes
#print len(seeds)
n_iteration = 500
if not os.path.exists(file_name):
diamond.DIAMOnD(network, seeds, n_iteration, alpha = 1, outfile = file_name)
f = open(file_name)
f.readline()
genes = []
for line in f:
rank, geneid = line.strip("\n").split()
genes.append(geneid)
f.close()
if not os.path.exists(file_name + ".coverage"):
f_out = open(file_name + ".coverage", 'w')
n = float(len(seeds))
component = network.subgraph(seeds)
#component = max(networkx.connected_components(component), key=len)
components = max(network_utilities.get_connected_components(network, False), key=len)
f_out.write("%s %f\n" % ("0", len(component & seeds)/n))
for i, gene in enumerate(genes):
rank = i + 1
component = network.subgraph(genes[:rank] + list(seeds))
#component = max(networkx.connected_components(component), key=len)
components = max(network_utilities.get_connected_components(network, False), key=len)
f_out.write("%s %f\n" % (rank, len(component & seeds)/n))
f_out.close()
return genes, nodes
def get_modules_of_graph(sub_graph,
module_detection_type,
output_file,
inflation=1.7):
if module_detection_type == "connected":
import network_utilities
modules = network_utilities.get_connected_components(
sub_graph, return_as_graph_list=True)
elif module_detection_type == "mcl":
from os import system
f = open(output_file + ".mcl", 'w')
nodes = set()
for node1, node2, data in sub_graph.edges(data=True):
nodes.add(node1)
nodes.add(node2)
if 'w' in data:
data = str(data['w'])
else:
data = "-"
f.write("%s\t%s\t%s\n" % (node1, node2, data))
for node in sub_graph.nodes():
if node not in nodes:
f.write("%s\n" % node)
f.close()
# Optimum inflation parameter was 1.7-1.8 in a recent comparison paper
system("mcl %s --abc -I %f -o %s 2>> %s" %
(output_file + ".mcl", inflation, output_file,
output_file + ".err"))
modules = get_modules_from_file(output_file)
else:
raise ValueError("Unrecognized module detection type")
#print len(modules), map(len, modules)
return modules
def get_modules_of_graph(sub_graph, module_detection_type, output_file, inflation=1.7):
if module_detection_type == "connected":
import network_utilities
modules = network_utilities.get_connected_components(sub_graph, return_as_graph_list=True)
elif module_detection_type == "mcl":
from os import system
f = open(output_file + ".mcl", 'w')
nodes = set()
for node1, node2, data in sub_graph.edges(data=True):
nodes.add(node1)
nodes.add(node2)
if 'w' in data:
data = str(data['w'])
else:
data = "-"
f.write("%s\t%s\t%s\n" % (node1, node2, data))
for node in sub_graph.nodes():
if node not in nodes:
f.write("%s\n" % node)
f.close()
# Optimum inflation parameter was 1.7-1.8 in a recent comparison paper
system("mcl %s --abc -I %f -o %s 2>> %s" % (output_file + ".mcl", inflation, output_file, output_file + ".err"))
modules = get_modules_from_file(output_file)
else:
raise ValueError("Unrecognized module detection type")
#print len(modules), map(len, modules)
return modules
def get_network(network_file, only_lcc):
network = network_utilities.create_network_from_sif_file(network_file, use_edge_data = False, delim = None, include_unconnected=True)
#print len(network.nodes()), len(network.edges())
if only_lcc:
components = network_utilities.get_connected_components(network, False)
network = network_utilities.get_subgraph(network, components[0])
#print len(network.nodes()), len(network.edges())
return network
def calculate_lcc_significance(network, nodes, nodes_random=None, bins=None, n_random=1000, min_bin_size=100, seed=452456):
if bins is None and nodes_random is None:
bins = network_utilities.get_degree_binning(network, min_bin_size)
if nodes_random is None:
nodes_random = get_random_nodes(nodes, network, bins = bins, n_random = n_random, min_bin_size = min_bin_size, seed = seed)
network_sub = network.subgraph(nodes)
component_nodes = network_utilities.get_connected_components(network_sub, False)[0]
d = len(component_nodes)
values = numpy.empty(len(nodes_random))
for i, nodes in enumerate(nodes_random):
network_sub = network.subgraph(nodes)
component_nodes = network_utilities.get_connected_components(network_sub, False)[0]
values[i] = len(component_nodes)
m, s = numpy.mean(values), numpy.std(values)
if s == 0:
z = 0.0
else:
z = (d - m) / s
return d, z, (m, s)
def get_network(network_file, only_lcc):
network = network_utilities.create_network_from_sif_file(network_file, use_edge_data = False, delim = None, include_unconnected=True)
#print len(network.nodes()), len(network.edges())
if only_lcc:
components = network_utilities.get_connected_components(network, False)
network = network_utilities.get_subgraph(network, components[0])
#print len(network.nodes()), len(network.edges())
network_lcc_file = network_file + ".lcc"
if not os.path.exists(network_lcc_file ):
f = open(network_lcc_file, 'w')
for u,v in network.edges():
f.write("%s 1 %s\n" % (u, v))
f.close()
return network
def get_network(network_file, only_lcc):
network = network_utilities.create_network_from_sif_file(network_file, use_edge_data = False, delim = None, include_unconnected=True)
#print len(network.nodes()), len(network.edges())
if only_lcc and not network_file.endswith(".lcc"):
print "Shrinking network to its LCC", len(network.nodes()), len(network.edges())
components = network_utilities.get_connected_components(network, False)
network = network_utilities.get_subgraph(network, components[0])
print "Final shape:", len(network.nodes()), len(network.edges())
#print len(network.nodes()), len(network.edges())
network_lcc_file = network_file + ".lcc"
if not os.path.exists(network_lcc_file ):
f = open(network_lcc_file, 'w')
for u,v in network.edges():
f.write("%s 1 %s\n" % (u, v))
f.close()
return network
def get_diseasome_genes(diseasome_file, nodes=None, network=None):
"""
If nodes is not None, keep only nodes in the network
If network is not None, keep only LCC
"""
disease_to_genes = {}
disease_to_category = {}
for line in open(diseasome_file):
words = line.strip("\n").split("\t")
disease = words[1].strip('"')
category = words[0]
genes = set(words[2:])
if nodes is not None:
genes &= nodes
if len(genes) == 0:
continue
if network is not None:
network_sub = network.subgraph(genes)
genes = network_utilities.get_connected_components(network_sub, False)[0]
disease_to_genes[disease] = genes
disease_to_category[disease] = category
return disease_to_genes, disease_to_category
