From 185bd346189b4bd6029ae26b9d52e29bce6ef4b0 Mon Sep 17 00:00:00 2001
From: AndiMajore <andi.majore@googlemail.com>
Date: Wed, 27 Jul 2022 19:18:29 +0200
Subject: [PATCH] all algorithms working for symbol space
---
tasks/closeness_centrality.py | 20 +++++++-----
tasks/network_proximity.py | 2 +-
tasks/quick_task.py | 49 +++++++----------------------
tasks/trust_rank.py | 4 +--
tasks/util/read_graph_tool_graph.py | 25 +++++++--------
tasks/util/scores_to_results.py | 2 --
6 files changed, 39 insertions(+), 63 deletions(-)
diff --git a/tasks/closeness_centrality.py b/tasks/closeness_centrality.py
index 7ae2772..1c306ee 100755
--- a/tasks/closeness_centrality.py
+++ b/tasks/closeness_centrality.py
@@ -1,3 +1,4 @@
+import numpy as np
from tasks.util.read_graph_tool_graph import read_graph_tool_graph
from tasks.util.scores_to_results import scores_to_results
from tasks.util.edge_weights import edge_weights
@@ -117,7 +118,6 @@ def closeness_centrality(task_hook: TaskHook):
# Reasonable default: False.
# Has no effect unless trust_rank.py is used for ranking drugs.
include_indirect_drugs = task_hook.parameters.get("include_indirect_drugs", False)
-
# Type: bool
# Semantics: Sepcifies whether should be included in the analysis when ranking drugs.
# Example: False.
@@ -164,7 +164,7 @@ def closeness_centrality(task_hook: TaskHook):
pdi_dataset = task_hook.parameters.get("pdi_dataset")
- search_target = task_hook.parameters.get("target", "drug-target")
+ search_target = task_hook.parameters.get("target", "drug")
filterPaths = task_hook.parameters.get("filter_paths", True)
@@ -173,11 +173,12 @@ def closeness_centrality(task_hook: TaskHook):
id_space = task_hook.parameters["config"].get("identifier", "symbol")
+ node_name_attribute = "internal_id"
+
filename = f"{id_space}_{ppi_dataset['name']}-{pdi_dataset['name']}"
if ppi_dataset['licenced'] or pdi_dataset['licenced']:
filename += "_licenced"
filename = os.path.join(task_hook.data_directory, filename + ".gt")
- # g, seed_ids, viral_protein_ids, drug_ids = read_graph_tool_graph(file_path, seeds, datasets, ignored_edge_types, max_deg, ignore_non_seed_baits, include_indirect_drugs, include_non_approved_drugs)
g, seed_ids, drug_ids = read_graph_tool_graph(filename, seeds, id_space, max_deg, include_indirect_drugs, include_non_approved_drugs, search_target)
task_hook.set_progress(1 / 4.0, "Computing edge weights.")
weights = edge_weights(g, hub_penalty)
@@ -189,12 +190,15 @@ def closeness_centrality(task_hook: TaskHook):
# Call graph-tool to compute TrustRank.
task_hook.set_progress(2 / 4.0, "Computing shortest path closeness centralities.")
all_dists = []
+ # score_nodes = drug_ids if search_target == 'drug' else seed_ids
for node in seed_ids:
- all_dists.append(gtt.shortest_distance(g, node, weights=weights))
- scores = len(seeds) / sum([dists.get_array() for dists in all_dists])
-
+ dists = gtt.shortest_distance(g, node, weights=weights).get_array()
+ dists[dists == np.inf] = 99999999999
+ all_dists.append(dists+1)
+
+ scores = len(seed_ids) / (sum([dists for dists in all_dists]))
+
# Compute and return the results.
- task_hook.set_progress(3 / 4.0, "Formating results.")
- # task_hook.set_results(scores_to_results(strain_or_drugs, result_size, g, seed_ids, viral_protein_ids, drug_ids, scores))
+ task_hook.set_progress(3 / 4.0, "Formatting results.")
task_hook.set_results(scores_to_results(search_target, result_size, g, seed_ids, drug_ids, scores, ppi_dataset, pdi_dataset, filterPaths))
diff --git a/tasks/network_proximity.py b/tasks/network_proximity.py
index 716c7a3..c07beee 100755
--- a/tasks/network_proximity.py
+++ b/tasks/network_proximity.py
@@ -79,7 +79,7 @@ def network_proximity(task_hook: TaskHook):
filter_paths = task_hook.parameters.get("filter_paths", True)
- node_name_attribute = "drugstone_id" # nodes in the input network which is created from RepoTrialDB have primaryDomainId as name attribute
+ node_name_attribute = "internal_id" # nodes in the input network which is created from RepoTrialDB have primaryDomainId as name attribute
# Set number of threads if OpenMP support is enabled.
if gt.openmp_enabled():
gt.openmp_set_num_threads(num_threads)
diff --git a/tasks/quick_task.py b/tasks/quick_task.py
index 21f4cbf..899674d 100755
--- a/tasks/quick_task.py
+++ b/tasks/quick_task.py
@@ -1,25 +1,14 @@
from tasks.task_hook import TaskHook
-
-def infer_node_type(node): # TODO: This needs to be improved
- if len(node) == 6 or len(node) == 10:
- return 'protein'
- # if node.startswith('DB'):
- # return 'drug'
- # return 'virus'
- if node.startswith('DB'):
- return 'drug'
- return 'protein'
-
-
def quick_task(task_hook: TaskHook):
- def run_closeness(parameters):
+ def run_closeness(parameters, network):
from .closeness_centrality import closeness_centrality
def closeness_progress(progress, status):
task_hook.set_progress(2 / 3 + 1 / 3 * progress, status)
def closeness_set_result(result):
+ result["network"]["edges"].extend(network["edges"])
task_hook.set_results(result)
# Prepare intermediate hook
@@ -31,25 +20,6 @@ def quick_task(task_hook: TaskHook):
# Run closeness centrality
closeness_centrality(closeness_task_hook)
- def run_trust_rank(parameters, seeds):
- from .trust_rank import trust_rank
-
- def progress(progress, status):
- task_hook.set_progress(2 / 3 + 1 / 3 * progress, status)
-
- def set_result(result):
- task_hook.set_results(result)
-
- parameters.update({
- "seeds": seeds,
- "result_size": 20,
- "include_non_approved_drugs": True,
- "include_indirect_drugs": False,
- "target":"drug"
- })
-
- tr_task_hook = TaskHook(parameters, task_hook.data_directory, progress, set_result)
- trust_rank(tr_task_hook)
def run_multi_steiner(parameters):
from .multi_steiner import multi_steiner
@@ -60,15 +30,20 @@ def quick_task(task_hook: TaskHook):
def ms_set_result(result):
node_attributes = result.get("node_attributes", {})
node_types = node_attributes.get("node_types", {})
- # seeds = [seed for seed in result["network"]["nodes"] if node_types.get(seed) == 'host' or
- # (not node_types.get(seed) and infer_node_type(seed) == 'host')]
- seeds = [seed for seed in result["network"]["nodes"] if node_types.get(seed) == 'protein' or
- (not node_types.get(seed) and infer_node_type(seed) == 'protein')]
+ seeds = [seed for seed in result["network"]["nodes"] if node_types.get(seed) == 'protein']
+
if len(seeds) == 0:
task_hook.set_results({"network": {"nodes": [], "edges": []}})
return
- run_trust_rank(parameters, seeds)
+ parameters.update({
+ "seeds": seeds,
+ "result_size": 10,
+ "hub_penalty": 1,
+ "target": "drug",
+ "include_non_approved_drugs": True
+ })
+ run_closeness(parameters, result["network"])
parameters["num_trees"] = 1
parameters["hub_penalty"] = 1
diff --git a/tasks/trust_rank.py b/tasks/trust_rank.py
index ec75373..6922fa6 100755
--- a/tasks/trust_rank.py
+++ b/tasks/trust_rank.py
@@ -218,8 +218,8 @@ def trust_rank(task_hook: TaskHook):
trust = g.new_vertex_property("double")
trust.a[seed_ids] = 1.0 / len(seed_ids)
scores = gtc.pagerank(g, damping=damping_factor, pers=trust, weight=weights)
-
# Compute and return the results.
task_hook.set_progress(3 / 4.0, "Formating results.")
# Convert results to useful output and save it
- task_hook.set_results(scores_to_results(search_target, result_size, g, seed_ids, drug_ids, scores, ppi_dataset, pdi_dataset, filter_paths))
+ results = scores_to_results(search_target, result_size, g, seed_ids, drug_ids, scores, ppi_dataset, pdi_dataset, filter_paths)
+ task_hook.set_results(results)
diff --git a/tasks/util/read_graph_tool_graph.py b/tasks/util/read_graph_tool_graph.py
index 35a89c4..2db32cf 100755
--- a/tasks/util/read_graph_tool_graph.py
+++ b/tasks/util/read_graph_tool_graph.py
@@ -51,7 +51,6 @@ def read_graph_tool_graph(file_path, seeds, id_space, max_deg, include_indirect_
# Remove all unconnected nodes TODO probably already skip when creating .gt files
if g.vertex(node).out_degree() == 0 and target == 'drug':
deleted_nodes.append(node)
- # if not g.vertex_properties["name"][node] in set(seeds) and g.vertex(node).out_degree() > max_deg:
elif not g.vertex_properties[node_name_attribute][node] in set(seeds) and (
g.vertex(node).out_degree() > max_deg):
deleted_nodes.append(node)
@@ -74,8 +73,8 @@ def read_graph_tool_graph(file_path, seeds, id_space, max_deg, include_indirect_
if include_non_approved_drugs:
drug_ids.append(node)
else:
- drug_groups = g.vertex_properties["status"][node].split(', ')
- if "approved" in drug_groups:
+ # drug_groups = g.vertex_properties["status"][node].split(', ')
+ if "approved" in g.vertex_properties["status"][node]:
drug_ids.append(node)
# Delete edges that should be ignored or are not contained in the selected dataset.
@@ -111,26 +110,26 @@ def read_graph_tool_graph(file_path, seeds, id_space, max_deg, include_indirect_
if indir_drug and int(edge.target()) in drug_ids:
drug_ids.remove(int(edge.target()))
- elif g.vertex_properties["type"][
- edge.source()] == d_type and edge.source() not in direct_drugs or edge.target() not in seed_ids:
+ elif g.vertex_properties["type"][edge.source()] == d_type and \
+ edge.source() not in direct_drugs or edge.target() not in seed_ids:
indir_drug = edge.source() not in direct_drugs
not_seed = edge.target() not in seed_ids
if indir_drug or not_seed:
deleted_edges.append(edge)
if indir_drug and int(edge.source()) in drug_ids:
drug_ids.remove(int(edge.source()))
- else:
- deleted_edges.append(edge)
+ # else:
+ # deleted_edges.append(edge)
g.set_fast_edge_removal(fast=True)
for edge in deleted_edges:
g.remove_edge(edge)
g.set_fast_edge_removal(fast=False)
- vertices = 0
- for _ in g.vertices():
- vertices += 1
- edges = 0
- for _ in g.edges():
- edges += 1
+ # vertices = 0
+ # for _ in g.vertices():
+ # vertices += 1
+ # edges = 0
+ # for _ in g.edges():
+ # edges += 1
# Return the graph and the indices of the seed_ids and the seeds.
return g, list(seed_ids.keys()), drug_ids
diff --git a/tasks/util/scores_to_results.py b/tasks/util/scores_to_results.py
index e3b679e..4db5e69 100755
--- a/tasks/util/scores_to_results.py
+++ b/tasks/util/scores_to_results.py
@@ -16,8 +16,6 @@ def scores_to_results(
r"""Transforms the scores to the required result format."""
node_name_attribute = "internal_id" # nodes in the input network which is created from RepoTrialDB have primaryDomainId as name attribute
- candidates = []
- # if strain_or_drugs == "drugs":
if target == "drug":
candidates = [(node, scores[node]) for node in drug_ids if scores[node] > 0]
else:
--
GitLab