Add files via upload

synthesized_data/all_kinase.py

@@ -37,12 +37,12 @@ def load_sequence_dict(sequence_df, sequence_field, phospho_field):
 kiba_mol = set(molListkiba)
 pdb.set_trace()
 '''
-dfm = pd.read_csv('../metz_data/restructured.csv', header = 0, index_col=False)
+dfm = pd.read_csv('../metz_data/restructured_unique.csv', header = 0, index_col=False)
 dfd = pd.read_csv('../davis_data/restructured.csv', header = 0, index_col=False)
-dfk = pd.read_csv('../KIBA_data/restructured.csv', header = 0, index_col=False)
-# dfk = dfk.head(200)
-# dfm = dfm.head(200)
-# dfd = dfd.head(200)
+dfk = pd.read_csv('../KIBA_data/restructured_unique.csv', header = 0, index_col=False)
+dfk = dfk.head(30000)
+dfm = dfm.head(12000)
+dfd = dfd.head(12000)
 #dfm = dfm.sort_values(by=['proteinName', 'smiles'])
 #dfd = dfd.sort_values(by=['proteinName', 'smiles'])
 kiba_sequence_df = pd.read_csv('../KIBA_data/prot_desc.csv', header=0, 
@@ -62,7 +62,7 @@ def load_sequence_dict(sequence_df, sequence_field, phospho_field):
 df_dicts = {dict_names[0]: dfk, dict_names[1]: dfm, dict_names[2]: dfd}
 #pdb.set_trace()
 
-invalid_val = -100
+invalid_val = np.nan
 davis_value = []
 metz_value = []
 kiba_value = []
@@ -163,7 +163,7 @@ def load_sequence_dict(sequence_df, sequence_field, phospho_field):
 #interaction_bin = (interactions >= 7.6) * 1
 
 shuffled = np.random.permutation(range(len(kiba_value)))
-with open('davis-metz-kiba.csv', 'w', newline='') as csvfile:
+with open('davis_metz_kiba2.csv', 'w', newline='') as csvfile:
   fieldnames = ['davis', 'metz', 'kiba', 'smiles', 'proteinName', 'protein_dataset']
   writer = csv.DictWriter(csvfile, fieldnames = fieldnames)
   writer.writeheader()
@@ -175,12 +175,12 @@ def load_sequence_dict(sequence_df, sequence_field, phospho_field):
     davis_val = davis_value[index]    
     metz_val = metz_value[index]      
     kiba_val = kiba_value[index] 
-    davis_val = '' if davis_val == invalid_val else davis_val    
-    metz_val = '' if metz_val == invalid_val else metz_val    
-    kiba_val = '' if kiba_val == invalid_val else kiba_val
+    davis_val = '' if davis_val != davis_val else davis_val    
+    metz_val = '' if metz_val != metz_val else metz_val    
+    kiba_val = '' if kiba_val != kiba_val else kiba_val
     writer.writerow({'davis': davis_val, 'metz': metz_val, 'kiba': kiba_val,
       'smiles': cmpd_smiles[index], 'proteinName': protein_name[index],
       'protein_dataset': protein_origin[index]})
 
 time_finish = time.time()
-print("time used: ", time_finish - time_start)
+print("time used: ", time_finish - time_start)

synthesized_data/create_davis_metz.py

@@ -0,0 +1,136 @@
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+from __future__ import unicode_literals
+
+import numpy as np
+import tensorflow as tf
+import pandas as pd
+import argparse
+import os
+import sys
+import pwd
+import pdb
+import csv
+import re
+import math
+import time
+
+def load_sequence_dict(sequence_df, sequence_field, phospho_field):
+  prot_desc_dict = {}
+  for row in sequence_df.itertuples():
+    sequence = row[sequence_field]
+    phosphorylated = row[phospho_field]
+    prot_desc_dict[row[0]] = [phosphorylated, sequence]
+  return prot_desc_dict
+
+'''
+dfdavis = pd.read_csv('../davis_data/Bio_results.csv', header = 2, index_col=0, usecols=range(3, 76))
+molListdavis = list(dfdavis)
+dfmetz = pd.read_csv('../metz_data/Metz_interaction.csv', header = 0, index_col=0, usecols=range(2, 16))
+molListmetz = list(dfmetz.index)
+molListmetz = [molName for molName in molListmetz if molName == molName]
+davis_mol = set(molListdavis)
+metz_mol = set(molListmetz)
+pdb.set_trace()
+'''
+dfm = pd.read_csv('../metz_data/restructured.csv', header = 0, index_col=False)
+dfd = pd.read_csv('../davis_data/restructured.csv', header = 0, index_col=False)
+#dfm = dfm.head(10)
+#dfd = dfd.head(200)
+#dfm = dfm.sort_values(by=['proteinName', 'smiles'])
+#dfd = dfd.sort_values(by=['proteinName', 'smiles'])
+
+metz_sequence_df = pd.read_csv('../metz_data/prot_desc_Metz.csv', header=0, 
+  index_col=0, usecols=range(0, 6))
+davis_sequence_df = pd.read_csv('../davis_data/prot_desc.csv', header=0, 
+  index_col=0, usecols=range(0, 3))
+time_start = time.time()
+metz_sequence_dict = load_sequence_dict(metz_sequence_df, 4, 5)
+davis_sequence_dict = load_sequence_dict(davis_sequence_df, 1, 2)
+#pdb.set_trace()
+
+invalid_val = -100
+davis_value = []
+metz_value = []
+protein_name = []
+protein_origin = []
+cmpd_smiles = []
+davis_leftout = [True]*len(dfd)
+prot_correspondence_dict = {}
+cmpd_correspondence_dict = {}
+
+for i in range(len(dfm)):
+  metz_prot = dfm.iloc[i]['proteinName']
+  metz_cmpd = dfm.iloc[i]['smiles']
+  metz_intxn = dfm.iloc[i]['interaction_value']
+  prot_identity = metz_sequence_dict[metz_prot]
+  if metz_prot not in prot_correspondence_dict:
+    prot_correspondence = [davis_sequence_dict[a] == prot_identity for a 
+      in dfd.iloc[:]['proteinName']]
+    prot_correspondence_dict[metz_prot] = prot_correspondence
+  else:
+    prot_correspondence = prot_correspondence_dict[metz_prot]
+
+  if metz_cmpd not in cmpd_correspondence_dict:
+    cmpd_correspondence = [metz_cmpd == b for b in dfd.iloc[:]['smiles']]
+    cmpd_correspondence_dict[metz_cmpd] = cmpd_correspondence
+  else:
+    cmpd_correspondence = cmpd_correspondence_dict[metz_cmpd]
+
+  exact_correspondence = [a and b for a, b in zip(prot_correspondence, 
+    cmpd_correspondence)]
+  #if sum(prot_correspondence) > 0:
+  #  pdb.set_trace()
+  if sum(exact_correspondence):
+    # There are exact correspondences.
+    position = np.where(exact_correspondence)[0]
+    assert len(position) == 1
+    for pos in position:
+      davis_intxn = dfd.iloc[pos]['interaction_value']
+      davis_leftout[pos] = False
+  else:
+    davis_intxn = invalid_val
+  davis_value.append(davis_intxn)
+  metz_value.append(metz_intxn)
+  protein_name.append(metz_prot)
+  protein_origin.append('metz')
+  cmpd_smiles.append(metz_cmpd)
+
+#pdb.set_trace()
+remaining_davis_entries = np.where(davis_leftout)[0]
+for entry in remaining_davis_entries:
+  davis_prot = dfd.iloc[entry]['proteinName']
+  davis_cmpd = dfd.iloc[entry]['smiles']
+  prot_identity = davis_sequence_dict[davis_prot]
+  davis_value.append(dfd.iloc[entry]['interaction_value'])
+  metz_value.append(invalid_val)
+  protein_name.append(davis_prot)
+  protein_origin.append('davis')
+  cmpd_smiles.append(davis_cmpd)
+  davis_leftout[entry] = False
+
+metz_value = np.array(metz_value)
+davis_value = np.array(davis_value)
+#interaction_bin = (interactions >= 7.6) * 1
+
+shuffled = np.random.permutation(range(len(metz_value)))
+with open('davis-metz.csv', 'w', newline='') as csvfile:
+  fieldnames = ['davis', 'metz', 'smiles', 'proteinName', 'protein_dataset']
+  writer = csv.DictWriter(csvfile, fieldnames = fieldnames)
+  writer.writeheader()
+  for i in range(len(davis_value)):    
+    # will start writing rows.    
+    #intxn_bin = interaction_bin[i][j]
+    index = shuffled[i]
+    davis_val = davis_value[index]    
+    metz_val = metz_value[index]      
+
+    davis_val = '' if davis_val == invalid_val else davis_val    
+    metz_val = '' if metz_val == invalid_val else metz_val
+    writer.writerow({'davis': davis_val, 'metz': metz_val,
+      'smiles': cmpd_smiles[index], 'proteinName': protein_name[index],
+      'protein_dataset': protein_origin[index]})
+
+time_finish = time.time()
+print("time used: ", time_finish-time_start)