import os, sys, json
from clustergram import clustergram
from orm_utils import *
sys.path.append('../../maayanlab_utils')
from fileIO import read_df
def find_name_for_id(uid):
projection = {'_id':False, 'limma':False, 'fold_changes':False, 'chdir': False}
dbs = DBSignature(uid,projection)
return dbs.name + '|' + dbs.meta['geo_id']
# mat, ids, _ = read_df('../signed_jaccard_subset_unique_entries_831x831.txt')
# mat, ids, _ = read_df('../signed_jaccard_subset_unique_entries_519x519.txt')
mat, ids, _ = read_df('../signed_jaccard_subset_unique_entries_259x259.txt')
names = map(find_name_for_id, ids)
json_data = clustergram(mat, names, names,
row_linkage='average', col_linkage='average',
row_pdist='cosine', col_pdist='cosine')
json.dump(json_data, open('/Library/WebServer/Documents/d3_clustergram/signed_jaccard_subset_clustergram_259.json', 'wb'))
## convert repFPKM table to MySQL database
import os, sys
import MySQLdb
import numpy as np
sys.path.append('C:\Users\Zichen\Documents\\bitbucket\maayanlab_utils')
from fileIO import read_df
os.chdir('D:\Zichen_Projects\Rendl_RNAseq')
mat, gene_tids, samples = read_df('repFpkmMatrix_allGenes.txt')
print mat.shape, len(gene_tids), len(samples)
d_sig = {}
with open ('signatures.txt') as f:
signatures = next(f).split('\t')
for line in f:
sl = line.split('\t')
for gene, sig in zip(sl, signatures):
if gene != '' and gene.strip() != '':
d_sig[gene.strip()] = sig.strip()
# from pprint import pprint
# pprint(d_sig)
conn = MySQLdb.connect(host='localhost',user='******', passwd='',db='hairgel')
cur = conn.cursor()
ii = 0
for row, gene_tid in zip(mat, gene_tids):
genes = gene_tid.split('|')[0].split(',')
## transfer drug tables
transfer_table('drugs_lincs', DrugLINCS, session)
transfer_table('drugs_drugbank', DrugDrugbank, session)
transfer_table('drugs_stitch', DrugStitch, session)
# p_vals from the latest prediction
## get AUC for each side_effect
# aucs = np.loadtxt(HOME + '/Documents/Zichen_Projects/drug_se_prediction/ExtraTrees100_RLogit_GO + MACCS_per-label_AUC_n794x1053.txt').mean(axis=1)
aucs = np.loadtxt(HOME + '/Documents/Zichen_Projects/drug_se_prediction/ET100_balanced_BRLogit2_auto_GO + MACCS_per-label_AUC-S3folds_n794x1053_top_50_features.txt').mean(axis=1)
print len(aucs)
print np.percentile(aucs, 75)
print aucs[aucs > 0.7].shape
mat, pert_ids, se_names = read_df(HOME + '/Documents/Zichen_Projects/drug_se_prediction/PTs_RF1000_proba_df_n20338x1053.txt')
print mat.shape
se_names = np.array(se_names)
d_se_auc = dict(zip(se_names, aucs))
d_umls_soc = pickle.load(open(HOME+'/Documents/Zichen_Projects/drug_se_prediction/d_umls_soc.p', 'rb'))
d_soc_color = pickle.load(open(HOME+'/Documents/Zichen_Projects/drug_se_prediction/d_soc_color.p', 'rb'))
## add soc table from d_soc_color
for soc, color in d_soc_color.items():
instance = get_or_create(session, SOC, name=soc, color=color)
## transfer side_effects and add AUROC
conn = MySQLdb.connect(host='localhost',user='******', passwd='',db='maaya0_SEP')
cur = conn.cursor()
query = """SELECT * FROM `%s`""" %'side_effects'
transfer_table('drugs_stitch', DrugStitch, session)
# p_vals from the latest prediction
## get AUC for each side_effect
# aucs = np.loadtxt(HOME + '/Documents/Zichen_Projects/drug_se_prediction/ExtraTrees100_RLogit_GO + MACCS_per-label_AUC_n794x1053.txt').mean(axis=1)
aucs = np.loadtxt(
HOME +
'/Documents/Zichen_Projects/drug_se_prediction/ET100_balanced_BRLogit2_auto_GO + MACCS_per-label_AUC-S3folds_n794x1053_top_50_features.txt'
).mean(axis=1)
print len(aucs)
print np.percentile(aucs, 75)
print aucs[aucs > 0.7].shape
mat, pert_ids, se_names = read_df(
HOME +
'/Documents/Zichen_Projects/drug_se_prediction/PTs_RF1000_proba_df_n20338x1053.txt'
)
print mat.shape
se_names = np.array(se_names)
d_se_auc = dict(zip(se_names, aucs))
d_umls_soc = pickle.load(
open(HOME + '/Documents/Zichen_Projects/drug_se_prediction/d_umls_soc.p',
'rb'))
d_soc_color = pickle.load(
open(HOME + '/Documents/Zichen_Projects/drug_se_prediction/d_soc_color.p',
'rb'))
## add soc table from d_soc_color
for soc, color in d_soc_color.items():
instance = get_or_create(session, SOC, name=soc, color=color)
def find_name_for_id(uid):
projection = {
'_id': False,
'limma': False,
'fold_changes': False,
'chdir': False
}
dbs = DBSignature(uid, projection)
return dbs.name + '|' + dbs.meta['geo_id']
# mat, ids, _ = read_df('../signed_jaccard_subset_unique_entries_831x831.txt')
# mat, ids, _ = read_df('../signed_jaccard_subset_unique_entries_519x519.txt')
mat, ids, _ = read_df('../signed_jaccard_subset_unique_entries_259x259.txt')
names = map(find_name_for_id, ids)
json_data = clustergram(mat,
names,
names,
row_linkage='average',
col_linkage='average',
row_pdist='cosine',
col_pdist='cosine')
json.dump(
json_data,
open(
'/Library/WebServer/Documents/d3_clustergram/signed_jaccard_subset_clustergram_259.json',