def nct_tagging(index_name, process_ids, stopwords, umls, pos, nprocs=1):
# open the clinical trail ids file to process
nct_ids = []
for line in open(process_ids, 'rb'):
nct_ids.append(line.strip())
# Check if index exists
index = es_index.ElasticSearch_Index(index_name)
# Create a new sentence index
sent_index_name = index_name + '_sent'
sent_index = es_index.ElasticSearch_Index(sent_index_name)
sent_index.open_index()
sent_index.add_field('ec_raw_text')
sent_index.add_field('ec_tags_umls', term_vector=True)
# Get clinical
# process each clinical trial and store to XML file
log.info('processing clinical trials')
procs = []
chunksize = int(math.ceil(len(nct_ids) / float(nprocs)))
for i in xrange(nprocs):
p = Process(target=_worker,
args=(nct_ids[chunksize * i:chunksize * (i + 1)],
index_name, sent_index_name, stopwords, umls, pos,
(i + 1)))
procs.append(p)
p.start()
for p in procs:
p.join()
def _worker(nct, index_name, host, port_no, stopwords, umls, pos, npr):
index = es_index.ElasticSearch_Index(index_name, host=host, port=port_no)
tagger = Tagger(5, stopwords, umls, pos)
# Iterate over NCT trials
for i in xrange(1, len(nct) + 1):
nctid = nct[i - 1]
# if i == 3:
#break
# if nctid != 'NCT00000331':
# continue
if i % 500 == 0:
log.info(' --- core %d: processed %d documents' % (npr, i))
# Get document from the Elastic Search Index
doc = index.get_trail(nctid)
if not doc.has_key('ec_raw_text'):
continue
ec = doc['ec_raw_text']
if ec is None:
continue
(pec, jpec) = tagger.process(ec)
dictlist = []
for key, value in jpec.iteritems():
for i in range(value):
dictlist.append(key)
doc['ec_tags_umls'] = dictlist
print nctid, dictlist
# Index the new document
index.index_trial(nctid, doc)
def _worker(nct, index_name, sent_index_name, stopwords, umls, pos, npr):
index = es_index.ElasticSearch_Index(index_name)
sent_index = es_index.ElasticSearch_Index(sent_index_name)
for i in xrange(1, len(nct) + 1):
nctid = nct[i - 1]
print nctid
if i % 500 == 0:
log.info(' --- core %d: indexed %d documents' % (npr, i))
doc = index.get_trail(nctid)
ec = doc['ec_raw_text']
if ec == None:
continue
sent_id = 1
for it in ec:
sent = ec[it].split(' - ')
for s in sent:
tags = {}
proc = TextProcesser(s, 5, stopwords, umls, pos)
proc.process()
for pp in proc.ptxt:
freq = tags.setdefault(pp, 0)
tags[pp] = freq + 1
if len(tags) == 0:
continue
dictlist = []
for key, value in substring_filtering(tags, 1).iteritems():
temp = [key, value]
dictlist.append(temp)
# Index the tags
sent_data = {}
sent_data['id'] = nctid + '_' + str(sent_id)
sent_data['ec_raw_text'] = s.strip()
sent_data['ec_tags_umls'] = dictlist
sent_index.index_trial(nctid + '_' + str(sent_id), sent_data)
sent_id += 1
def _worker(nct, data_path, index_name, host, port_no, npr):
index = es_index.ElasticSearch_Index(index_name, host=host, port=port_no)
parser = ctgov_parser.ClinicalTrial_Parser(data_path)
for i in xrange(1, len(nct) + 1):
nctid = nct[i - 1]
# if i == 3:
# break
if i % 500 == 0:
log.info(' --- core %d: indexed %d documents' % (npr, i))
trail_doc = parser.parse(nctid)
index.index_trial(nctid, trail_doc)
def update_ES_index(nct_disease, index_name, host, port_no):
index = es_index.ElasticSearch_Index(index_name, host=host, port=port_no)
# Iterate over NCT trials
for nctid in nct_disease.keys():
# Get document from the Elastic Search Index
doc = index.get_trail(nctid)
if doc is False or not doc.has_key('ec_raw_text'):
continue
ec = doc['ec_raw_text']
if ec is None:
continue
doc['disease'] = nct_disease[nctid]
index.index_trial(nctid, doc)
def nct_index(din,
index_name,
host='localhost',
port_no=9200,
nprocs=1,
settings_file=None):
# open the clinical trail ids file and load to a list
log.info('opening file -- trial_ids.txt')
nct_ids = []
for line in open(din + '/trial_ids.txt', 'rb'):
nct_ids.append(line.strip())
# Check directories
trials_din = din + '/trials_xml/'
if not os.path.exists(trials_din):
log.error('trials_xml directory does not exists in %s \n' % din)
exit(0)
# Create index using the provided settings file
index = es_index.ElasticSearch_Index(index_name, host=host, port=port_no)
if settings_file == None:
index.open_index()
else:
index.open_index(settings_file)
index.add_field('conditions', term_vector=False)
index.add_field('study_type', term_vector=False)
# process each clinical trial and store to XML file
log.info('processing clinical trials')
procs = []
chunksize = int(math.ceil(len(nct_ids) / float(nprocs)))
for i in xrange(nprocs):
p = Process(target=_worker,
args=(nct_ids[chunksize * i:chunksize * (i + 1)],
trials_din, index_name, host, port_no, (i + 1)))
procs.append(p)
p.start()
for p in procs:
p.join()
def build_graph(ec_index_name, host, port, neo_url, umls):
# Initialize Graph
graph_db = neo4j.GraphDatabaseService(neo_url)
# Create CDE index
cde_index = graph_db.get_or_create_index(neo4j.Node, 'cde-index')
disease_index = graph_db.get_or_create_index(neo4j.Node, 'disease-index')
# Initialize ElasticSearch Index
index = es_index.ElasticSearch_Index(ec_index_name, host, port)
# Get all unique conditions
conditions = index.get_unique_terms('conditions', min_docs=2)
log.info(' Found %d conditions' % len(conditions))
# Add Diseases (conditions) with properties to GraphDB
batch = neo4j.WriteBatch(graph_db)
for disease, freq in conditions.iteritems():
tmp = batch.create(node(name=disease, trail_with_disease=freq))
batch.set_labels(tmp, 'Disease')
batch.add_to_index(neo4j.Node, disease_index, 'name', disease, tmp)
batch.run()
log.info(' Successfully added %d conditions' % len(conditions))
# Get all unique CDEs
# cdes = index.get_unique_terms('ec_tags_umls', min_docs=3)
#print 'cdes = ', len(cdes)
cdes = index.get_unique_terms('ec_tags_umls', min_docs=5)
log.info(' Found %d CDEs' % len(cdes))
# Add CDEs with properties to GraphDB
batch = neo4j.WriteBatch(graph_db)
for cde, freq in cdes.iteritems():
stype = get_semantic_types(cde, umls)
tmp = batch.create(node(name=cde, num_trail=freq,
semantic_types=stype))
batch.set_labels(tmp, 'CDE')
batch.add_to_index(neo4j.Node, cde_index, 'name', cde, tmp)
batch.run()
log.info('Successfully added %d CDEs' % len(cdes))
# For each disease-CDE pair get counts
rel_count = 0
i = 0
for cde in cdes.keys():
i += 1
batch = neo4j.WriteBatch(graph_db)
cde_node = graph_db.get_indexed_node('cde-index', 'name', cde)
for disease in conditions.keys():
freq = get_disease_cde_count(index, disease, cde)
if freq <= 2:
continue
rel_count += 1
disease_node = graph_db.get_indexed_node('disease-index', 'name',
disease)
batch.create(rel(cde_node, 'related_to', disease_node, freq=freq))
if i % 50 == 0:
log.info('Rel count at %d -- %d out of %d CDEs seen' %
(rel_count, i, len(cdes)))
batch.run()
log.info('Successfully added %d relationships' % rel_count)
