def save_pkl(self, fn):
db = {
'decipher': self.decipher,
'orpha': self.orpha,
'omim': self.omim
}
save_object(db, fn)
def gen_patient_embeddings(source,
EXP_ID,
exp_id,
enriched='no',
n_same_time=None,
exp_variable=None):
start = time.time()
if n_same_time:
patients_phenotypes = \
'./_data/patients/'+source+'_patients_phenotype_'+n_same_time+'.csv'
patient_embeddings = \
'./_data/patients/'+source+'_patient_embeddings_'+n_same_time+'.pkl'
else:
patients_phenotypes = \
'./_data/patients/'+source+'_patients_phenotype.csv'
patient_embeddings = \
'./_data/patients/'+source+'_patient_embeddings.pkl'
with open(patients_phenotypes) as csv_file:
patient_sims = csv.reader(csv_file)
hpo = Hpo()
hpo_vectors = HpoVecs(enriched,
EXP_ID,
exp_id,
exp_variable=exp_variable).vecs
patients = {}
for line in patient_sims:
patients[line[0]] = \
compute_embedding_average(line[1:], hpo, hpo_vectors)
save_object(patients, patient_embeddings)
print(time.time() - start)
def save_pkl(self, filename):
db = {}
db['decipher'] = self.decipher
db['decipher_desc'] = self.decipher_desc
db['orpha'] = self.orpha
db['orpha_desc'] = self.orpha_desc
db['omim'] = self.omim
db['omim_desc'] = self.omim_desc
save_object(db, filename)
def learn_embeddings(self, walks):
'''
Learn embeddings by optimizing the Skipgram objective using SGD.
'''
# walks = [map(str, walk) for walk in walks] #py2
walks = [list(map(str, walk)) for walk in walks]
save_object(walks, './_data/walks/walks_hpo_orpha.pkl')
model = Word2Vec(walks,
size=self.dimensions,
window=self.window_size,
min_count=0,
sg=1,
workers=self.workers,
iter=self.iter)
# model.save_word2vec_format(args.output) #deprecated
model.wv.save_word2vec_format(self.output)
return
def gen_mapping_objects(source, n_same_time=None):
start = time.time()
if n_same_time:
with open('./_emu/emu-' + source + '_' + n_same_time +
'.json') as json_file:
patient_sims = json.load(json_file)
f = open('_data/patients/' + source + '_patients_phenotype_' +
n_same_time + '.csv',
'w',
newline='')
writer = csv.writer(f)
patients_conditions = {}
fn = '_data/patients/' + source + '_patients_disease_' + n_same_time + '.pkl'
else:
with open('./_emu/emu-' + source + '.json') as json_file:
patient_sims = json.load(json_file)
f = open('_data/patients/' + source + '_patients_phenotype.csv',
'w',
newline='')
writer = csv.writer(f)
patients_conditions = {}
fn = '_data/patients/' + source + '_patients_disease.pkl'
i = 0
for disease in patient_sims:
for phenotype in patient_sims[disease]['sims']:
identifier = 'P' + str(i)
id_arr = [identifier]
for symptom in phenotype:
id_arr.append(symptom)
writer.writerow(id_arr)
patients_conditions[identifier] = disease
i += 1
f.close()
save_object(patients_conditions, fn)
print(time.time() - start)
walk_length = 50
iterations = 3
conds = 200
noise_ptgs = [.15, .3, .45, .6]
patients_per_cond = 2
lamb = 4
enriched_embeddings = 'no'
# print('Generating embeddings...')
# start_time_symptom_embeddings = time.time()
# genEmbeddings(input='_data/graph/hp-obo.edgelist', output='_data/emb/hp-obo_'+EXP_ID+'_'+str(exp_int)+'.emb', walk_length=walk_length, iter=iterations)
# amount_time_symptom_embeddings = time.time()-start_time_symptom_embeddings
rows = []
metadata_list = []
save_object(metadata_list, '_data/results/experiment_number' + EXP_ID + '.pkl')
sim_names = {
'cos_sim', 'jaccard_best_avg', 'resnik_best_avg', 'lin_best_avg',
'jc_best_avg'
}
fig, ax = plt.subplots(2, 2)
i = 0
for noise_ptg in noise_ptgs:
aucs = {
'cos_sim': [],
'jaccard_best_avg': [],
'resnik_best_avg': [],
'lin_best_avg': [],
'jc_best_avg': []
}
def save_pkl(self, fn):
save_object(self.hpos, fn)