def network_layout(gmt_fn, outfn=None):
## make a Graph object and write to gml for Gephi to
## do the layout
d_gmt = read_gmt(gmt_fn)
d_gmt_filt = {}
for term, genes in d_gmt.items():
if len(genes) >= 5:
d_gmt_filt[term] = genes
d_gmt = d_gmt_filt
print 'number of terms:', len(d_gmt)
umls_ids_kept = d_gmt.keys()
adj_matrix = jaccard_matrix(d_gmt)
m = adj_matrix > 0.2
# degrees = adj_matrix.sum(axis=0)
adj_matrix = adj_matrix * m.astype(int)
G = nx.from_numpy_matrix(adj_matrix)
print 'G: ',G.number_of_edges(), G.number_of_nodes()
for i in range(adj_matrix.shape[0]):
# G.node[i]['size'] = degrees[i]
# G.node[i]['size'] = len(d_gmt[umls_ids_kept[i]])
G.node[i]['size'] = G.degree(i)
G.node[i]['id'] = umls_ids_kept[i]
if outfn is not None:
nx.write_gml(G, outfn)
return G
def network_layout(gmt_fn, outfn=None):
## make a Graph object and write to gml for Gephi to
## do the layout
d_gmt = read_gmt(gmt_fn)
d_gmt_filt = {}
for term, genes in d_gmt.items():
if len(genes) >= 5:
d_gmt_filt[term] = genes
d_gmt = d_gmt_filt
print 'number of terms:', len(d_gmt)
umls_ids_kept = d_gmt.keys()
adj_matrix = jaccard_matrix(d_gmt)
m = adj_matrix > 0.2
# degrees = adj_matrix.sum(axis=0)
adj_matrix = adj_matrix * m.astype(int)
G = nx.from_numpy_matrix(adj_matrix)
print 'G: ', G.number_of_edges(), G.number_of_nodes()
for i in range(adj_matrix.shape[0]):
# G.node[i]['size'] = degrees[i]
# G.node[i]['size'] = len(d_gmt[umls_ids_kept[i]])
G.node[i]['size'] = G.degree(i)
G.node[i]['id'] = umls_ids_kept[i]
if outfn is not None:
nx.write_gml(G, outfn)
return G
def make_directed_json_graph_soc(gmt_fn, d_id_name, d_id_category, d_category_color, outfn=None):
# make directed graph based on SOC - PT
d_gmt = read_gmt(gmt_fn)
d_gmt_filt = {}
for term, genes in d_gmt.items():
if len(genes) >= 5:
d_gmt_filt[term] = genes
d_gmt = d_gmt_filt
print 'number of terms:', len(d_gmt)
umls_ids_kept = d_gmt.keys()
adj_matrix = jaccard_matrix(d_gmt)
m = adj_matrix > 0.2
adj_matrix = adj_matrix * m.astype(int)
Gu = nx.from_numpy_matrix(adj_matrix) # undirected Graph, to get size
G = nx.DiGraph()
for i in range(len(umls_ids_kept)):
umls_id = umls_ids_kept[i]
name = d_id_name[umls_id]
category = d_id_category[umls_id]
color = d_category_color[category]
G.add_edge('root', category)
G.add_edge(category, umls_id)
G.node[umls_id]['size'] = Gu.degree(i)
G.node[umls_id]['label'] = name
G.node[umls_id]['color'] = color
print G.number_of_nodes(), G.number_of_edges()
graph_data = json_graph.tree_data(G,root='root')
json.dump(graph_data, open(outfn, 'wb'))
return
def make_directed_json_graph(gmt_fn,
d_id_name,
d_id_category,
d_category_color,
outfn=None):
# perform HC and make a directed graph and write to json
# for pack visualization
d_gmt = read_gmt(gmt_fn)
d_gmt_filt = {}
for term, genes in d_gmt.items():
if len(genes) >= 5:
d_gmt_filt[term] = genes
d_gmt = d_gmt_filt
print 'number of terms:', len(d_gmt)
umls_ids_kept = d_gmt.keys()
adj_matrix = jaccard_matrix(d_gmt)
hc = AgglomerativeClustering(n_clusters=10)
hc.fit(adj_matrix)
m = adj_matrix > 0.2
adj_matrix = adj_matrix * m.astype(int)
Gu = nx.from_numpy_matrix(adj_matrix) # undirected Graph, to get size
G = nx.DiGraph()
print adj_matrix.shape, len(umls_ids_kept)
for i in range(adj_matrix.shape[0]):
cluster_label = hc.labels_[i]
umls_id = umls_ids_kept[i]
name = d_id_name[umls_id]
G.add_edge('root', cluster_label)
G.add_edge(cluster_label, umls_id)
G.node[umls_id]['size'] = Gu.degree(i)
G.node[umls_id]['label'] = name
category = d_id_category[umls_id]
color = d_category_color[category]
G.node[umls_id]['color'] = color
print G.number_of_nodes(), G.number_of_edges()
graph_data = json_graph.tree_data(G, root='root')
json.dump(graph_data, open(outfn, 'wb'))
return
def make_directed_json_graph(gmt_fn, d_id_name, d_id_category, d_category_color, outfn=None):
# perform HC and make a directed graph and write to json
# for pack visualization
d_gmt = read_gmt(gmt_fn)
d_gmt_filt = {}
for term, genes in d_gmt.items():
if len(genes) >= 5:
d_gmt_filt[term] = genes
d_gmt = d_gmt_filt
print 'number of terms:', len(d_gmt)
umls_ids_kept = d_gmt.keys()
adj_matrix = jaccard_matrix(d_gmt)
hc = AgglomerativeClustering(n_clusters=10)
hc.fit(adj_matrix)
m = adj_matrix > 0.2
adj_matrix = adj_matrix * m.astype(int)
Gu = nx.from_numpy_matrix(adj_matrix) # undirected Graph, to get size
G = nx.DiGraph()
print adj_matrix.shape, len(umls_ids_kept)
for i in range(adj_matrix.shape[0]):
cluster_label = hc.labels_[i]
umls_id = umls_ids_kept[i]
name = d_id_name[umls_id]
G.add_edge('root', cluster_label)
G.add_edge(cluster_label, umls_id)
G.node[umls_id]['size'] = Gu.degree(i)
G.node[umls_id]['label'] = name
category = d_id_category[umls_id]
color = d_category_color[category]
G.node[umls_id]['color'] = color
print G.number_of_nodes(), G.number_of_edges()
graph_data = json_graph.tree_data(G,root='root')
json.dump(graph_data, open(outfn, 'wb'))
return
def make_directed_json_graph_soc(gmt_fn,
d_id_name,
d_id_category,
d_category_color,
outfn=None):
# make directed graph based on SOC - PT
d_gmt = read_gmt(gmt_fn)
d_gmt_filt = {}
for term, genes in d_gmt.items():
if len(genes) >= 5:
d_gmt_filt[term] = genes
d_gmt = d_gmt_filt
print 'number of terms:', len(d_gmt)
umls_ids_kept = d_gmt.keys()
adj_matrix = jaccard_matrix(d_gmt)
m = adj_matrix > 0.2
adj_matrix = adj_matrix * m.astype(int)
Gu = nx.from_numpy_matrix(adj_matrix) # undirected Graph, to get size
G = nx.DiGraph()
for i in range(len(umls_ids_kept)):
umls_id = umls_ids_kept[i]
name = d_id_name[umls_id]
category = d_id_category[umls_id]
color = d_category_color[category]
G.add_edge('root', category)
G.add_edge(category, umls_id)
G.node[umls_id]['size'] = Gu.degree(i)
G.node[umls_id]['label'] = name
G.node[umls_id]['color'] = color
print G.number_of_nodes(), G.number_of_edges()
graph_data = json_graph.tree_data(G, root='root')
json.dump(graph_data, open(outfn, 'wb'))
return
## for side effects
# GMT_FN = HOME+'/Documents/Zichen_Projects/drug_se_prediction/RF1000_GOtCS_AUC_0.7_proba_0.6_prediction_only.gmt'
GMT_FN = HOME+'/Documents/Zichen_Projects/drug_se_prediction/ET100_GOtCS_AUC_0.76_proba_0.75.gmt'
GML_FN = HOME+'/Documents/Zichen_Projects/drug_se_prediction/side_effect_network.gml'
## for drugs
# GMT_FN = HOME+'/Documents/Zichen_Projects/drug_se_prediction/RF1000_GOtCS_AUC_0.7_proba_0.6_prediction_only_flipped.gmt'
# GML_FN = HOME+'/Documents/Zichen_Projects/drug_se_prediction/drug_network.gml'
CSV_FN = GML_FN.replace('.gml', '.csv')
JSON_FN = CSV_FN.replace('.csv', '.json')
## retrieve meta data about SE
d_umls_pt = mysqlTable2dict('sep', 'side_effects', 1, 2)
d_pt_umls = mysqlTable2dict('sep', 'side_effects', 2, 1)
d_soc_pt = read_gmt(HOME+'/Documents/bitbucket/pertid2trainingset/Y_matrix_no_mfc/SOC_to_pt.gmt')
print len(d_soc_pt)
d_umls_soc = {}
for soc, pts in d_soc_pt.items():
for pt in pts:
umls = d_pt_umls[pt]
if umls is not None:
if umls not in d_umls_soc:
d_umls_soc[umls] = soc
# else:
# d_umls_soc[umls].append(soc) # one PT may have multiple SOCs
# for pt in d_umls_soc:
# if len(d_umls_soc[pt]) != 1:
# print pt, d_umls_soc[pt]
## for side effects
# GMT_FN = HOME+'/Documents/Zichen_Projects/drug_se_prediction/RF1000_GOtCS_AUC_0.7_proba_0.6_prediction_only.gmt'
GMT_FN = HOME + '/Documents/Zichen_Projects/drug_se_prediction/ET100_GOtCS_AUC_0.76_proba_0.75.gmt'
GML_FN = HOME + '/Documents/Zichen_Projects/drug_se_prediction/side_effect_network.gml'
## for drugs
# GMT_FN = HOME+'/Documents/Zichen_Projects/drug_se_prediction/RF1000_GOtCS_AUC_0.7_proba_0.6_prediction_only_flipped.gmt'
# GML_FN = HOME+'/Documents/Zichen_Projects/drug_se_prediction/drug_network.gml'
CSV_FN = GML_FN.replace('.gml', '.csv')
JSON_FN = CSV_FN.replace('.csv', '.json')
## retrieve meta data about SE
d_umls_pt = mysqlTable2dict('sep', 'side_effects', 1, 2)
d_pt_umls = mysqlTable2dict('sep', 'side_effects', 2, 1)
d_soc_pt = read_gmt(
HOME +
'/Documents/bitbucket/pertid2trainingset/Y_matrix_no_mfc/SOC_to_pt.gmt')
print len(d_soc_pt)
d_umls_soc = {}
for soc, pts in d_soc_pt.items():
for pt in pts:
umls = d_pt_umls[pt]
if umls is not None:
if umls not in d_umls_soc:
d_umls_soc[umls] = soc
# else:
# d_umls_soc[umls].append(soc) # one PT may have multiple SOCs
# for pt in d_umls_soc:
# if len(d_umls_soc[pt]) != 1:
# print pt, d_umls_soc[pt]