pk_file = open('tiedie_kernel.pkl', 'wb')
pickle.dump(diffuser, pk_file)
pk_file.close()
# Check to make sure the node universe matches the
k_labels = diffuser.getLabels()
if len(network_nodes) != len(k_labels) or len(
network_nodes.intersection(k_labels)) != len(k_labels):
sys.stderr.write(
"Error: the universe of gene/node labels in the network file doesn't match the supplied kernel file!\n"
)
sys.exit(1)
sys.stderr.write("Diffusing Heats...\n")
# Separately perform diffusion for each input set
up_heats_diffused = diffuser.diffuse(up_heats, reverse=False)
down_heats_diffused = diffuser.diffuse(down_heats, reverse=True)
# Extract a subnetwork solution from the diffused heats, the set of nodes found, the Relevance score for this network solution
# and the heat scores for all linker genes
subnet_soln, subnet_soln_nodes, alpha_score, linker_scores = extractSubnetwork(
up_heats, down_heats, up_heats_diffused, down_heats_diffused, size_control,
opts.alpha)
#
# Generate linker stats and output
#
out_degrees = getOutDegrees(subnet_soln)
sys.stderr.write("Writing network node stats to " + output_folder +
"/node.stats\n")
out_file = output_folder + "/node.stats"
#
# Diffusion Step:
# Load the heat diffusion kernel and perform a kernel-multiply, or alternatively use supplied
# page-rank diffused vectors
#
print "Parsing Network File..."
network = parseNet(opts.network)
if opts.pagerank:
diffuser = PPrDiffuser(network)
else:
print "Loading Heat Diffusion Kernel.."
diffuser = Kernel(opts.kernel)
print "Diffusing Heats..."
up_heats_diffused = diffuser.diffuse(up_heats, reverse=False)
down_heats_diffused = diffuser.diffuse(down_heats, reverse=True)
if opts.size_select:
# Search a number of network sizes to get a range with the highest significance.
# Somewhat slow.
search_space = []
start, end = opts.size_select.split(":")
start = int(start)
end = int(end)
for size in range(1, 20):
search_space.append(float(size)/20)
for size in range(11, 20):
search_space.append(float(size)/10)
for size in range(1, 15):
