From b6378426311739b2aad0fb18b2cebe51e7ef0da9 Mon Sep 17 00:00:00 2001
From: AndiMajore <andi.majore@googlemail.com>
Date: Tue, 26 Jul 2022 16:25:58 +0200
Subject: [PATCH] fixed must and kpm; removed prints
Former-commit-id: 67306ed60f13b0702c2c6ed812f0fc18140505f0 [formerly 97b6706ddca64fc407e2639e1c182a6bc5ae78b9]
Former-commit-id: 7f2eb37d1159e20af4a4bb18c7425a53c4a57d7d
---
tasks/keypathwayminer_task.py | 23 ++++++++++++++----
tasks/multi_steiner.py | 36 ++++-------------------------
tasks/trust_rank.py | 1 -
tasks/util/read_graph_tool_graph.py | 2 --
tasks/util/steiner_tree.py | 6 -----
5 files changed, 23 insertions(+), 45 deletions(-)
diff --git a/tasks/keypathwayminer_task.py b/tasks/keypathwayminer_task.py
index 54080de..639d589 100755
--- a/tasks/keypathwayminer_task.py
+++ b/tasks/keypathwayminer_task.py
@@ -1,5 +1,6 @@
import base64
import datetime
+import itertools
import json
import random
import string
@@ -199,18 +200,32 @@ def kpm_task(task_hook: TaskHook):
# Remapping everything from UniProt Accession numbers to internal IDs
result_nodes = Protein.objects.filter(uniprot_code__in=network["nodes"])
node_map = {}
+ node_map_for_edges = {}
+
for node in result_nodes:
- node_map[node.uniprot_code] = node.id
- network["nodes"] = list(map(lambda uniprot: "p" + str(node_map[uniprot]), network["nodes"]))
+ node_map_for_edges[node.uniprot_code] = node.id
+ if id_space == 'symbol':
+ node_map[node.uniprot_code] = [node.gene]
+ if id_space == 'entrez':
+ node_map[node.uniprot_code] = [node.entrez]
+ if id_space == 'uniprot':
+ node_map[node.uniprot_code] = [node.uniprot_code]
+ if id_space == 'ensembl':
+ node_map[node.uniprot_code] = [ensg.name for ensg in EnsemblGene.objects.filter(protein_id=node.id)]
+
+ flat_map = lambda f, xs: [y for ys in xs for y in f(ys)]
+
+ network["nodes"] = flat_map(lambda uniprot: node_map[uniprot], network["nodes"])
network["edges"] = list(map(
- lambda uniprot_edge: {"from": "p" + str(node_map[uniprot_edge["from"]]),
- "to": "p" + str(node_map[uniprot_edge["to"]])},
+ lambda uniprot_edge: {"from": "p" + str(node_map_for_edges[uniprot_edge["from"]]),
+ "to": "p" + str(node_map_for_edges[uniprot_edge["to"]])},
network["edges"]))
node_types = {node: "protein" for node in network["nodes"]}
is_seed = {node: node in set(map(lambda p: "p"+str(p),protein_backend_ids)) for node in network["nodes"]}
result_dict = {
"network": network,
+ "target_nodes":[node for node in network["nodes"] if node not in task_hook.seeds],
"node_attributes": {"node_types": node_types, "is_seed": is_seed}
}
task_hook.set_results(results=result_dict)
diff --git a/tasks/multi_steiner.py b/tasks/multi_steiner.py
index f361168..cc479ad 100755
--- a/tasks/multi_steiner.py
+++ b/tasks/multi_steiner.py
@@ -113,74 +113,48 @@ def multi_steiner(task_hook: TaskHook):
filename += "_licenced"
filename = os.path.join(task_hook.data_directory, filename + ".gt")
g, seed_ids, _ = read_graph_tool_graph(filename, seeds, id_space, max_deg, target=search_target)
- # seed_map = {g.vertex_properties["name"][node]: node for node in seed_ids}
seed_map = {g.vertex_properties[node_name_attribute][node]: node for node in seed_ids}
task_hook.set_progress(1 / (float(num_trees + 3)), "Computing edge weights.")
weights = edge_weights(g, hub_penalty)
# Find first steiner trees
seeds = list(filter(lambda s: s in seed_map, seeds))
- print(seeds)
- print(seed_ids)
task_hook.set_progress(2 / (float(num_trees + 3)), "Computing Steiner tree 1 of {}.".format(num_trees))
first_tree = steiner_tree(g, seeds, seed_map, weights, hub_penalty > 0)
num_found_trees = 1
tree_edges = []
for tree_edge in first_tree.edges():
- # source_name = first_tree.vertex_properties["name"][first_tree.vertex_index[tree_edge.source()]]
- # target_name = first_tree.vertex_properties["name"][first_tree.vertex_index[tree_edge.target()]]
- # tree_edges.append((gtu.find_vertex(g, prop=g.vertex_properties['name'], match=source_name)[0],gtu.find_vertex(g, prop=g.vertex_properties['name'], match=target_name)[0]))
source_name = first_tree.vertex_properties[node_name_attribute][first_tree.vertex_index[tree_edge.source()]]
target_name = first_tree.vertex_properties[node_name_attribute][first_tree.vertex_index[tree_edge.target()]]
tree_edges.append((gtu.find_vertex(g, prop=g.vertex_properties[node_name_attribute], match=source_name)[0], gtu.find_vertex(g, prop=g.vertex_properties[node_name_attribute], match=target_name)[0]))
cost_first_tree = sum([weights[g.edge(source, target)] for source, target in tree_edges])
- # returned_nodes = set(int(gtu.find_vertex(g, prop=g.vertex_properties['name'], match=first_tree.vertex_properties["name"][node])[0]) for node in range(first_tree.num_vertices()))
- print(f"Before gtu: Costs={cost_first_tree}")
returned_nodes = set(int(gtu.find_vertex(g, prop=g.vertex_properties[node_name_attribute], match=first_tree.vertex_properties[node_name_attribute][node])[0]) for node in range(first_tree.num_vertices()))
- print(f"After gtu: {returned_nodes}")
- print(num_trees)
if num_trees > 1:
- print("num_trees > 1")
is_bridge = find_bridges(g)
- print("found bridges")
edge_filter = g.new_edge_property("boolean", True)
- print("filtered edges")
found_new_tree = True
while len(tree_edges) > 0:
- print(f"Tree edges length: {len(tree_edges)}")
if found_new_tree:
task_hook.set_progress(float(num_found_trees + 2) / (float(num_trees + 3)), "Computing Steiner tree {} of {}.".format(num_found_trees + 1, num_trees))
found_new_tree = False
tree_edge = tree_edges.pop()
- print("1")
g_edge = g.edge(tree_edge[0], tree_edge[1])
if not is_bridge[g_edge]:
- print("2")
edge_filter[g_edge] = False
g.set_edge_filter(edge_filter)
next_tree = steiner_tree(g, seeds, seed_map, weights, hub_penalty > 0)
- print("3")
next_tree_edges = set()
for next_tree_edge in next_tree.edges():
- # source_name = next_tree.vertex_properties["name"][next_tree.vertex_index[next_tree_edge.source()]]
- # target_name = next_tree.vertex_properties["name"][next_tree.vertex_index[next_tree_edge.target()]]
- # next_tree_edges.add((gtu.find_vertex(g, prop=g.vertex_properties['name'], match=source_name)[0], gtu.find_vertex(g, prop=g.vertex_properties['name'], match=target_name)[0]))
- print("4")
source_name = next_tree.vertex_properties[node_name_attribute][next_tree.vertex_index[next_tree_edge.source()]]
target_name = next_tree.vertex_properties[node_name_attribute][next_tree.vertex_index[next_tree_edge.target()]]
next_tree_edges.add((gtu.find_vertex(g, prop=g.vertex_properties[node_name_attribute], match=source_name)[0],gtu.find_vertex(g, prop=g.vertex_properties[node_name_attribute], match=target_name)[0]))
cost_next_tree = sum([weights[g.edge(source, target)] for source, target in next_tree_edges])
if cost_next_tree <= cost_first_tree * ((100.0 + tolerance) / 100.0):
- print("5")
found_new_tree = True
num_found_trees += 1
for node in range(next_tree.num_vertices()):
- print("GTU again")
- # returned_nodes.add(int(gtu.find_vertex(g, prop=g.vertex_properties['name'], match=next_tree.vertex_properties["name"][node])[0]))
returned_nodes.add(int(gtu.find_vertex(g, prop=g.vertex_properties[node_name_attribute],match=next_tree.vertex_properties[node_name_attribute][node])[0]))
- print("GTU done")
removed_edges = []
- print("6")
for source, target in tree_edges:
if not ((source, target) in set(next_tree_edges)) or ((target, source) in set(next_tree_edges)):
removed_edges.append((source, target))
@@ -190,18 +164,15 @@ def multi_steiner(task_hook: TaskHook):
edge_filter[g_edge] = True
if num_found_trees >= num_trees:
break
-
task_hook.set_progress((float(num_trees + 2)) / (float(num_trees + 3)), "Formatting results")
returned_edges = []
for node in returned_nodes:
for neighbor in g.get_all_neighbors(node):
if int(neighbor) > node and int(neighbor) in returned_nodes:
returned_edges.append((node, int(neighbor)))
- # subgraph = {"nodes": [g.vertex_properties["name"][node] for node in returned_nodes],
- # "edges": [{"from": g.vertex_properties["name"][source], "to": g.vertex_properties["name"][target]} for source, target in returned_edges]}
- # node_types = {g.vertex_properties["name"][node]: g.vertex_properties["type"][node] for node in returned_nodes}
- # is_seed = {g.vertex_properties["name"][node]: node in set(seed_ids) for node in returned_nodes}
- subgraph = {"nodes": [g.vertex_properties[node_name_attribute][node] for node in returned_nodes],
+
+ accepted_nodes = [g.vertex_properties[node_name_attribute][node] for node in returned_nodes]
+ subgraph = {"nodes": accepted_nodes,
"edges": [{"from": g.vertex_properties[node_name_attribute][source], "to": g.vertex_properties[node_name_attribute][target]} for
source, target in returned_edges]}
node_types = {g.vertex_properties[node_name_attribute][node]: g.vertex_properties["type"][node] for node in returned_nodes}
@@ -209,6 +180,7 @@ def multi_steiner(task_hook: TaskHook):
task_hook.set_results({
"network": subgraph,
"node_attributes": {"node_types": node_types, "is_seed": is_seed},
+ "taget_nodes":accepted_nodes,
'gene_interaction_dataset': ppi_dataset,
'drug_interaction_dataset': pdi_dataset
})
diff --git a/tasks/trust_rank.py b/tasks/trust_rank.py
index fbcb5cc..ec75373 100755
--- a/tasks/trust_rank.py
+++ b/tasks/trust_rank.py
@@ -216,7 +216,6 @@ def trust_rank(task_hook: TaskHook):
# Call graph-tool to compute TrustRank.
task_hook.set_progress(2 / 4.0, "Computing TrustRank.")
trust = g.new_vertex_property("double")
- print(seed_ids)
trust.a[seed_ids] = 1.0 / len(seed_ids)
scores = gtc.pagerank(g, damping=damping_factor, pers=trust, weight=weights)
diff --git a/tasks/util/read_graph_tool_graph.py b/tasks/util/read_graph_tool_graph.py
index 92c1333..9901ae1 100755
--- a/tasks/util/read_graph_tool_graph.py
+++ b/tasks/util/read_graph_tool_graph.py
@@ -40,9 +40,7 @@ def read_graph_tool_graph(file_path, seeds, id_space, max_deg, include_indirect_
The graph indices for all drug nodes
"""
# Read the graph.
- print(f"loading {file_path} for {target}")
g = gt.load_graph(file_path)
- # g = gtt.extract_largest_component(gg, directed=False, prune=True) # this line is added since we need to work with the LCC of the graphs for all algorithms
# drug_protein = "DrugHasTarget"
d_type = "drug"
diff --git a/tasks/util/steiner_tree.py b/tasks/util/steiner_tree.py
index f91f647..b697243 100755
--- a/tasks/util/steiner_tree.py
+++ b/tasks/util/steiner_tree.py
@@ -4,7 +4,6 @@ import itertools as it
def steiner_tree(g, seeds, seed_map, weights, non_zero_hub_penalty):
-
node_name_attribute = "internal_id" # nodes in the input network which is created from RepoTrialDB have primaryDomainId as name attribute
mc = gt.Graph(directed=False)
eprop_dist = mc.new_edge_property("int")
@@ -80,10 +79,6 @@ def steiner_tree(g, seeds, seed_map, weights, non_zero_hub_penalty):
weights_g2[e_g2] = weights[e]
mst2 = gtt.min_spanning_tree(g2, root=None, tree_map=None, weights=weights_g2)
g2.set_edge_filter(mst2)
- # vw = gt.GraphView(g2, efilt=mst2)
- # g3 = Graph(vw, prune=True)
-
- # g3 = Graph(g22)
while True:
noneSteinerLeaves = []
@@ -94,7 +89,6 @@ def steiner_tree(g, seeds, seed_map, weights, non_zero_hub_penalty):
break
noneSteinerLeaves = reversed(sorted(noneSteinerLeaves))
for node in noneSteinerLeaves:
- # outarray = g3.get_out_edges(node)
g2.remove_edge(g2.edge(g2.vertex(node), g2.get_all_neighbors(node)[0]))
g2.remove_vertex(node)
--
GitLab