handling protein nodes now with multiple ids outside of main id space

Former-commit-id: 50c346aa

docker-django.env.dev

@@ -14,4 +14,4 @@ DJANGO_SETTINGS_MODULE=drugstone.settings
 CELERY_BROKER_URL=redis://redis:6379/0
 FLOWER_PORT=8888
 FLOWER_BASIC_AUTH=drugstone:test
-GT_THREADS=8
\ No newline at end of file
+GT_THREADS=2
\ No newline at end of file

drugstone/management/commands/make_graphs.py

@@ -95,7 +95,9 @@ def create_gt(params: Tuple) -> None:
     v_type = g.new_vertex_property("string")
     v_name = g.new_vertex_property("string")
     v_drugstone_id = g.new_vertex_property("string")
-    v_entrez = g.new_vertex_property("string")
+    v_has_symbol = g.new_vertex_property("bool")
+    v_has_entrez = g.new_vertex_property("bool")
+    v_has_ensembl = g.new_vertex_property("bool")
     v_expression = g.new_vertex_property("string")
 
     # for drugs
@@ -108,7 +110,9 @@ def create_gt(params: Tuple) -> None:
     g.vertex_properties["type"] = v_type
     g.vertex_properties["name"] = v_name
     g.vertex_properties["drugstone_id"] = v_drugstone_id
-    g.vertex_properties["entrez"] = v_entrez
+    g.vertex_properties["has_symbol"] = v_has_symbol
+    g.vertex_properties["has_entrez"] = v_has_entrez
+    g.vertex_properties["has_ensembl"] = v_has_ensembl
     g.vertex_properties["status"] = v_status
     g.vertex_properties["drug_id"] = v_drug_id
     g.vertex_properties["expression"] = v_expression
@@ -122,11 +126,16 @@ def create_gt(params: Tuple) -> None:
     print(f'loading nodes')
     # extend node data by cancer nodes, we create a normal node for each cancer node.
     # on reading the data, we decide which one to keep based on the user selected cancer types
+
+    has_ensembl_set = {node.protein_id for node in models.EnsemblGene.objects.all()}
+
     for node in models.Protein.objects.all():
         v = g.add_vertex()
         v_type[v] = 'protein'
         v_drugstone_id[v] = f"p{node.id}"
-
+        v_has_symbol[v] = len(node.gene) != 0
+        v_has_entrez[v] = len(node.entrez) != 0
+        v_has_ensembl[v] = node.id in has_ensembl_set
         vertices[node.id] = v
 
     print("done with nodes")
@@ -148,7 +157,6 @@ def create_gt(params: Tuple) -> None:
         e_type[e] = 'protein-protein'
     print("done with edges")
 
-
     print(f'loading drug_edges/{pdi_dataset}')
     for edge_raw in _internal_pdis(pdi_dataset):
         e = g.add_edge(drug_vertices[edge_raw.drug_id], vertices[edge_raw.protein_id])
@@ -161,7 +169,7 @@ def create_gt(params: Tuple) -> None:
         if vertex.out_degree() == 0:
             delete_vertices.add(vertex)
 
-    #remove unconnected drugs
+    # remove unconnected drugs
     for vertex in drug_vertices.values():
         if vertex.out_degree() == 0:
             delete_vertices.add(vertex)

drugstone/util/query_db.py

-from typing import List, Tuple, Set
+from collections import defaultdict
+from typing import List, Tuple, Set, OrderedDict
 from functools import reduce
 from django.db.models import Q
 from drugstone.models import Protein, EnsemblGene
@@ -31,7 +32,8 @@ def query_proteins_by_identifier(node_ids: Set[str], identifier: str) -> Tuple[L
         q_list = map(lambda n: Q(uniprot_code__iexact=n), node_ids)
     elif identifier == 'ensg':
         protein_attribute = 'ensg'
-        node_ids = map(lambda n: n.protein_id, EnsemblGene.objects.filter(reduce(lambda a,b: a|b, map(lambda n:Q(name__iexact=n),list(node_ids)))))
+        node_ids = map(lambda n: n.protein_id, EnsemblGene.objects.filter(
+            reduce(lambda a, b: a | b, map(lambda n: Q(name__iexact=n), list(node_ids)))))
         q_list = map(lambda n: Q(id=n), node_ids)
     elif identifier == 'entrez':
         protein_attribute = 'entrez'
@@ -40,9 +42,27 @@ def query_proteins_by_identifier(node_ids: Set[str], identifier: str) -> Tuple[L
         # node_ids is an empty list
         return [], protein_attribute
     q_list = reduce(lambda a, b: a | b, q_list)
-
     node_objects = Protein.objects.filter(q_list)
-    # serialize
-    nodes = ProteinSerializer(many=True).to_representation(node_objects)
+
+    nodes = list()
+
+    node_map = defaultdict(list)
+
+    for node in ProteinSerializer(many=True).to_representation(node_objects):
+        node_map[node.get(protein_attribute)].append(node)
+    for node_id, entries in node_map.items():
+        nodes.append(aggregate_nodes(entries))
 
     return nodes, protein_attribute
+
+
+def aggregate_nodes(nodes: List[OrderedDict]):
+    node = defaultdict(set)
+    for n in nodes:
+        for key, value in n.items():
+            if isinstance(value,list):
+                for e in value:
+                    node[key].add(e)
+            else:
+                node[key].add(value)
+    return {k: list(v) for k, v in node.items()}

drugstone/views.py

@@ -58,7 +58,7 @@ def get_pdis_ds(source, licenced):
 
 def get_drdis_ds(source, licenced):
     try:
-        ds =  models.PDisDataset.objects.filter(name__iexact=source, licenced=licenced).last()
+        ds = models.PDisDataset.objects.filter(name__iexact=source, licenced=licenced).last()
         ds.id
         return ds
     except:
@@ -128,7 +128,14 @@ def fetch_edges(request) -> Response:
         Response: List of edges which are objects with 'from' and to ' attribtues'
     """
     dataset = request.data.get('dataset', DEFAULTS['ppi'])
-    drugstone_ids = [node['drugstone_id'][1:] for node in request.data.get('nodes', '[]') if 'drugstone_id' in node]
+    drugstone_ids = set()
+    for node in request.data.get('nodes', '[]'):
+        if 'drugstone_id' in node:
+            if isinstance(node['drugstone_id'], list):
+                for id in node['drugstone_id']:
+                    drugstone_ids.add(id[1:])
+            else:
+                drugstone_ids.add(node['drugstone_id'])
     licenced = request.data.get('licenced', False)
     dataset_object = get_ppi_ds(dataset, licenced)
     interaction_objects = models.ProteinProteinInteraction.objects.filter(
@@ -168,16 +175,19 @@ def map_nodes(request) -> Response:
     nodes_mapped, id_key = query_proteins_by_identifier(node_ids, identifier)
 
     # change data structure to dict in order to be quicker when merging
-    if identifier == 'ensg':
-        # a protein might have multiple ensg-numbers, unpack these into single nodes
-        nodes_mapped_dict = {node_id: node for node in nodes_mapped for node_id in node[id_key]}
-    else:
-        nodes_mapped_dict = {node[id_key]: node for node in nodes_mapped}
+    # if identifier == 'ensg':
+    #     # a protein might have multiple ensg-numbers, unpack these into single nodes
+    #     nodes_mapped_dict = {node_id: node for node in nodes_mapped for node_id in node[id_key]}
+    # else:
+    nodes_mapped_dict = {node[id_key][0]: node for node in nodes_mapped}
     # merge fetched data with given data to avoid data loss
     for node in nodes:
+        node['drugstoneType'] = 'other'
         if node['id'] in nodes_mapped_dict:
             node.update(nodes_mapped_dict[node['id']])
+            node['drugstoneType'] = 'protein'
         node['id'] = id_map[node['id']]
+
     # set label to node identifier if label is unset, otherwise
     # return list of nodes updated nodes
     return Response(nodes)
@@ -489,6 +499,8 @@ def adjacent_disorders(request) -> Response:
         # serialize
         edges = DrugDisorderIndicationSerializer(many=True).to_representation(drdi_objects)
         disorders = DisorderSerializer(many=True).to_representation(disorders)
+    for d in disorders:
+        d['drugstone_type'] = 'disorder'
     return Response({
         'edges': edges,
         'disorders': disorders,
@@ -514,6 +526,9 @@ def adjacent_drugs(request) -> Response:
     # serialize
     pdis = ProteinDrugInteractionSerializer(many=True).to_representation(pdi_objects)
     drugs = DrugSerializer(many=True).to_representation(drugs)
+    for drug in drugs:
+        drug['drugstone_type'] = 'drug'
+
     return Response({
         'pdis': pdis,
         'drugs': drugs,

tasks/betweenness_centrality.py

@@ -170,13 +170,17 @@ def betweenness_centrality(task_hook: TaskHook):
 
     filterPaths = task_hook.parameters.get("filter_paths", True)
 
+    id_space = task_hook.parameters["config"].get("identifier","symbol")
+
+    print(id_space)
+
     # Parsing input file.
     task_hook.set_progress(0 / 3.0, "Parsing input.")
     filename = f"internal_{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, drug_ids = read_graph_tool_graph(
+    g, seed_ids, id_space, drug_ids = read_graph_tool_graph(
         filename,
         seeds,
         max_deg,

tasks/closeness_centrality.py

@@ -173,9 +173,12 @@ def closeness_centrality(task_hook: TaskHook):
     filename = f"internal_{ppi_dataset['name']}_{pdi_dataset['name']}"
     if ppi_dataset['licenced'] or pdi_dataset['licenced']:
         filename += "_licenced"
+
+    id_space = task_hook.parameters["config"].get("identifier", "symbol")
+
     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, max_deg, include_indirect_drugs, include_non_approved_drugs, search_target)
+    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) 
     

tasks/degree_centrality.py

@@ -153,9 +153,12 @@ def degree_centrality(task_hook: TaskHook):
     filename = f"internal_{ppi_dataset['name']}_{pdi_dataset['name']}"
     if ppi_dataset['licenced'] or pdi_dataset['licenced']:
         filename += "_licenced"
+
+    id_space = task_hook.parameters["config"].get("identifier", "symbol")
+
     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, False, include_non_approved_drugs)
-    g, seed_ids, drug_ids = read_graph_tool_graph(filename, seeds, max_deg, False, include_non_approved_drugs, search_target)
+    g, seed_ids, drug_ids = read_graph_tool_graph(filename, seeds, id_space, max_deg, False, include_non_approved_drugs, search_target)
     
     # Set number of threads if OpenMP support is enabled.
     if gt.openmp_enabled():

tasks/multi_steiner.py

@@ -105,11 +105,15 @@ def multi_steiner(task_hook: TaskHook):
     
     # Parsing input file.
     task_hook.set_progress(0 / (float(num_trees + 3)), "Parsing input.")
+
+    id_space = task_hook.parameters["config"].get("identifier", "symbol")
+
     filename = f"internal_{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, _ = read_graph_tool_graph(filename, seeds, max_deg, target=search_target)
+    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.")

tasks/network_proximity.py

@@ -86,12 +86,15 @@ def network_proximity(task_hook: TaskHook):
 
     # Parsing input file.
     task_hook.set_progress(0.0 / 8, "Parsing input.")
+
+    id_space = task_hook.parameters["config"].get("identifier", "symbol")
+
     filename = f"internal_{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, _, drug_ids = read_graph_tool_graph(file_path, seeds, "", "", max_deg, False, True, include_non_approved_drugs)
-    g, seed_ids, drug_ids = read_graph_tool_graph(filename, seeds, max_deg, True, include_non_approved_drugs, target=search_target)
+    g, seed_ids, drug_ids = read_graph_tool_graph(filename, seeds, id_space, max_deg, True, include_non_approved_drugs, target=search_target)
     # Computing edge weights.
     task_hook.set_progress(1.0 / 8, "Computing edge weights.")
     weights = edge_weights(g, hub_penalty)

tasks/trust_rank.py

@@ -198,11 +198,14 @@ def trust_rank(task_hook: TaskHook):
     
     # Parsing input file.
     task_hook.set_progress(0 / 4.0, "Parsing input.")
+
+    id_space = task_hook.parameters["config"].get("identifier", "symbol")
+
     filename = f"internal_{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, drug_ids = read_graph_tool_graph(filename, seeds, max_deg, include_indirect_drugs, include_non_approved_drugs, search_target)
+    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, inverse=True)
     

tasks/util/read_graph_tool_graph.py

@@ -3,7 +3,7 @@ import graph_tool.topology as gtt
 
 
 # def read_graph_tool_graph(file_path, seeds, datasets, ignored_edge_types, max_deg, ignore_non_seed_baits=False, include_indirect_drugs=False, include_non_approved_drugs=False):
-def read_graph_tool_graph(file_path, seeds, max_deg, include_indirect_drugs=False, include_non_approved_drugs=False,
+def read_graph_tool_graph(file_path, seeds,id_space, max_deg, include_indirect_drugs=False, include_non_approved_drugs=False,
                           target='drug'):
     r"""Reads a graph-tool graph from file.