PSEV/connect_all_patients.py at master · BaranziniLab/PSEV · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
"""Connects all overlapping concepts between EHRs and SPOKE
Usage:
  connect_all_patients.py [-i=NAME] [-s=NAME] [-o=NAME]
  connect_all_patients.py -h | --help

Options:
  -h --help     Show this screen.
  -i=NAME     Input directory [default: build/].
  -s=NAME     SPOKE directory [default: spoke_v_1/].
  -o=NAME     Output filename [default: all_patient_cohort].
"""
from docopt import docopt

import pickle as pic
import operator
import numpy as np
from collections import Counter
import time
import os
import pandas as pd

arguments = docopt(__doc__)

directory = arguments['-i']
spoke_directory = arguments['-s']
save_filename = arguments['-o']

node_list = np.load(spoke_directory+'spoke_node_list.npy', allow_pickle=False)
node_type_list = np.load(spoke_directory+'node_type_list.npy', allow_pickle=False)
#
doid_to_icd_filename = 'SPOKE_Disease_to_UCSF_EHR.tsv'
icd_9_cui_filename = 'ehr_cui_icd9.tsv'
icd_10_cui_filename = 'ehr_cui_icd10.tsv'
symptom_cui_and_mesh_filename = 'symptom_cui_and_mesh.tsv'
db_id_to_clarity_file = 'SPOKE_Compound_to_UCSF_EHR.tsv'
loinc_filename='distinct_cui_loinc.tsv'
path_filename='cui_to_loinc_path.tsv'
cui_cpt_filename='distinct_cui_cpt.tsv'
#
patient_filename = directory+'patient_cohort.tsv'
diagnoses_table_filename = directory+'EHR_DIAGNOSIS.tsv'
medication_table_filename = directory+'EHR_MEDICATION.tsv'
lab_table_filename = directory+'EHR_LABS.tsv'

def get_icd_cui_mesh_tables():
	disease_query_table = pd.read_table('constants/'+doid_to_icd_filename, sep='\t', header=0, dtype=str, index_col=0)
	icd9_table = pd.read_table('constants/'+icd_9_cui_filename, sep='\t', header=0)
	icd10_table = pd.read_table('constants/'+icd_10_cui_filename, sep='\t', header=0)
	mesh_table = pd.read_table('constants/'+symptom_cui_and_mesh_filename, sep='\t', header=0)
	cui_icd_table = pd.concat((icd9_table, icd10_table), ignore_index=True).rename(index=str, columns={"CODE": "EHR", "CUI":"SPOKE"}).drop_duplicates()
	mesh_cui_icd_table = pd.merge(mesh_table, cui_icd_table, left_on='CUI', right_on='SPOKE').drop(['CUI', 'SPOKE'], axis=1).rename(index=str, columns={"CODE":"SPOKE"}).drop_duplicates()
	icd_to_spoke = pd.concat((disease_query_table, cui_icd_table, mesh_cui_icd_table), ignore_index=True)
	icd_to_spoke = icd_to_spoke[icd_to_spoke['SPOKE'].isin(np.array(node_list, dtype=str))]
	return icd_to_spoke

def get_loinc_map_stats(path_df):
	id_and_type = path_df.drop(['EHR'], axis=1).drop_duplicates()
	print '# of LOINC mapped to SPOKE:', path_df.EHR.unique().shape[0]
	print 'Mapped SPOKE breakdown from LABS:', Counter(id_and_type.Node_Type.values)

def filter_transform_labs_to_spoke(path_len_filter=2):
	path_df = pd.read_table('constants/'+path_filename, sep='\t', header=0)
	cui_to_all_spoke = pd.read_table(spoke_directory+'All_SPOKE_CUI.tsv', sep='\t', header=0)
	cui_cpt_df = pd.read_table('constants/'+cui_cpt_filename, sep='\t', header=0)
	cui_loinc_df = pd.read_table('constants/'+loinc_filename, sep='\t', header=0)
	path_len_list, first_list, second_list = [], [], []
	for cui, path in path_df.values:
		path_list = path.split('->')
		path_len_list.append(len(path_list))
		first_list.append(path_list[0])
		second_list.append(path_list[1])
	path_df['Path_Len'] = np.array(path_len_list)
	path_df['First'] = np.array(first_list)
	path_df['Second'] = np.array(second_list)
	path_df = path_df.rename(index=str, columns={"CUI":"CUI_SPOKE"}).drop(['Path'], axis=1)
	path_df = path_df[path_df['Path_Len'] <= (path_len_filter+1)]
	cpt_path = path_df[path_df['Second'].isin(cui_cpt_df.CUI.unique())]
	path_df = pd.concat((path_df[path_df['Path_Len'] <= path_len_filter], cpt_path), ignore_index=True)
	path_df = pd.merge(path_df, cui_loinc_df, left_on='First', right_on='CUI').drop(['First', 'Second', 'Path_Len', 'CUI'], axis=1)
	path_df = pd.merge(cui_to_all_spoke, path_df, left_on='CUI', right_on='CUI_SPOKE').drop(['CUI', 'CUI_SPOKE'], axis=1).rename(index=str, columns={'Node_ID':'SPOKE', 'CODE':'EHR'}).drop_duplicates()
	get_loinc_map_stats(path_df)
	path_df = path_df.drop(['Node_Type'], axis=1).drop_duplicates()
	return path_df

def get_patient_to_spoke_df(icd_to_spoke, clarity_to_spoke, path_df):
	patient_diag_ehr = pd.read_table(diagnoses_table_filename, sep='\t', header=0)
	patient_diag_ehr = pd.concat((pd.merge(patient_diag_ehr.drop(['ICD10_Code'], axis=1), icd_to_spoke, left_on='ICD9_Code', right_on='EHR').drop(['EHR', 'ICD9_Code'], axis=1), pd.merge(patient_diag_ehr.drop(['ICD9_Code'], axis=1), icd_to_spoke, left_on='ICD10_Code', right_on='EHR').drop(['EHR', 'ICD10_Code'], axis=1)), ignore_index=True).drop_duplicates()
	#
	patient_med_ehr  = pd.read_table(medication_table_filename, sep='\t', header=0)
	patient_med_ehr['Medication_ID'] = np.array(patient_med_ehr['Medication_ID'], dtype=str)
	patient_med_ehr = pd.merge(patient_med_ehr, clarity_to_spoke, left_on='Medication_ID', right_on='EHR').drop(['EHR', 'Medication_ID'], axis=1).drop_duplicates()
	#
	patient_lab_ehr = pd.merge(pd.read_table(lab_table_filename, sep='\t', header=0), path_df, left_on='LOINC_Code', right_on='EHR').drop(['EHR', 'LOINC_Code'], axis=1).drop_duplicates()
	patient_lab_ehr = patient_lab_ehr[patient_lab_ehr['Lab_Result_Abnormal']=='Yes']
	patient_lab_ehr = patient_lab_ehr.drop(['Lab_Result_Abnormal'], axis=1)
	return pd.concat((patient_diag_ehr, patient_med_ehr, patient_lab_ehr), ignore_index=True)

def print_binary_matrix(patient_to_hetio_connect):
	output = open(directory+save_filename+'_binary_direct_hit_matrix', 'w')
	for row in patient_to_hetio_connect:
		output.write(' '.join(np.array(row, dtype=str))+'\n')
	output.close()

patient_list = np.array(pd.read_table(patient_filename, sep='\t', header=0)['Patient_ID'].values, dtype=str)
icd_to_spoke = get_icd_cui_mesh_tables()
clarity_to_spoke = pd.read_table('constants/'+db_id_to_clarity_file, sep='\t', header=0, dtype=str, index_col=0)
#
path_df = filter_transform_labs_to_spoke(path_len_filter=2)
#
patient_to_spoke = get_patient_to_spoke_df(icd_to_spoke, clarity_to_spoke, path_df)
patient_to_spoke['Patient_ID'] = np.array(patient_to_spoke['Patient_ID'].values, dtype=str)
patient_to_spoke['SPOKE'] = np.array(patient_to_spoke['SPOKE'].values, dtype=str)
patient_to_spoke = patient_to_spoke[patient_to_spoke['Patient_ID'].isin(patient_list)]
#
direct_hit_list = patient_to_spoke.SPOKE.unique()
bin_direct_hit = np.array([node in direct_hit_list for node in node_list])
direct_hit_list = np.array(node_list[bin_direct_hit], dtype=str)
print 'SEP STATS:', direct_hit_list.shape, Counter(node_type_list[bin_direct_hit])
np.save(directory+save_filename+'_direct_hit_list', direct_hit_list, allow_pickle=False)
np.save(directory+save_filename+'_spoke_node_list', node_list, allow_pickle=False)
#
dh_index_df = pd.DataFrame(data=np.array([direct_hit_list, np.arange(len(direct_hit_list))]).T, columns=['SPOKE', 'SEP_Index'])
patient_index_df = pd.DataFrame(data=np.array([patient_list, np.arange(len(patient_list))]).T, columns=['Patient_ID', 'Patient_Index'])
patient_to_spoke = pd.merge(pd.merge(patient_index_df, patient_to_spoke, on='Patient_ID'), dh_index_df, on='SPOKE')
#
patient_to_hetio_connect = np.zeros((len(patient_list), len(direct_hit_list)), dtype=int)
patient_to_hetio_connect[np.array(patient_to_spoke.Patient_Index.values, dtype=int) , np.array(patient_to_spoke.SEP_Index.values, dtype=int)] = 1
print_binary_matrix(patient_to_hetio_connect)
#