Merge branch 'attn_model' of github.com-orisenbazuru:uzh-dqbm-cmi/sid…

…e-effects into attn_model
uzh-dqbm-cmi · Jun 15, 2020 · 0d6dc4e · 0d6dc4e
2 parents addf902 + aa68123
commit 0d6dc4e
Show file tree

Hide file tree

Showing 16 changed files with 7,119 additions and 707 deletions.
diff --git a/...ns/backup/replicatedIndividualScores_max_epochs-20-batch_size-200-H1-300-H2-400-Copy1.csv b/...ns/backup/replicatedIndividualScores_max_epochs-20-batch_size-200-H1-300-H2-400-Copy1.csv
@@ -0,0 +1,9 @@
+Similarity,AUC,AUPR,F-measure,Recall,Precision
+chem,0.514,0.368,0.137,0.152,0.551
+target,0.634,0.527,0.494,0.453,0.561
+transporter,0.605,0.484,0.427,0.358,0.558
+enzyme,0.615,0.497,0.459,0.431,0.537
+pathway,0.618,0.505,0.467,0.429,0.529
+indication,0.681,0.596,0.567,0.555,0.581
+sideeffect,0.604,0.551,0.41,0.418,0.606
+offsideeffect,0.66,0.647,0.51,0.502,0.657
diff --git a/...ns/backup/replicatedIndividualScores_max_epochs-20-batch_size-200-H1-400-H2-300-Copy1.csv b/...ns/backup/replicatedIndividualScores_max_epochs-20-batch_size-200-H1-400-H2-300-Copy1.csv
@@ -0,0 +1,9 @@
+Similarity,AUC,AUPR,F-measure,Recall,Precision
+chem,0.512,0.385,0.111,0.094,0.426
+target,0.643,0.535,0.51,0.48,0.559
+transporter,0.594,0.474,0.4,0.328,0.556
+enzyme,0.614,0.491,0.452,0.418,0.537
+pathway,0.625,0.507,0.483,0.461,0.524
+indication,0.685,0.607,0.572,0.558,0.591
+sideeffect,0.618,0.581,0.443,0.436,0.612
+offsideeffect,0.638,0.655,0.465,0.46,0.669
diff --git a/...rocessed/runs/replicatedIndividualScores_max_epochs-20-batch_size-200-CNN-8736-256-64.csv b/...rocessed/runs/replicatedIndividualScores_max_epochs-20-batch_size-200-CNN-8736-256-64.csv
@@ -0,0 +1,9 @@
+Similarity,AUC,AUPR,F-measure,Recall,Precision
+chem,0.0,0.0,0.0,0.0,0.0
+target,0.0,0.0,0.0,0.0,0.0
+transporter,0.0,0.0,0.0,0.0,0.0
+enzyme,0.0,0.0,0.0,0.0,0.0
+pathway,0.0,0.0,0.0,0.0,0.0
+indication,0.0,0.0,0.0,0.0,0.0
+sideeffect,0.0,0.0,0.0,0.0,0.0
+offsideeffect,0.0,0.0,0.0,0.0,0.0
diff --git a/.../processed/runs/replicatedIndividualScores_max_epochs-20-batch_size-200-H1-300-H2-400.csv b/.../processed/runs/replicatedIndividualScores_max_epochs-20-batch_size-200-H1-300-H2-400.csv
@@ -1,9 +1,9 @@
 Similarity,AUC,AUPR,F-measure,Recall,Precision
-chem,0.0,0.0,0.0,0.0,0.0
-target,0.0,0.0,0.0,0.0,0.0
-transporter,0.0,0.0,0.0,0.0,0.0
-enzyme,0.0,0.0,0.0,0.0,0.0
-pathway,0.0,0.0,0.0,0.0,0.0
-indication,0.0,0.0,0.0,0.0,0.0
-sideeffect,0.0,0.0,0.0,0.0,0.0
-offsideeffect,0.0,0.0,0.0,0.0,0.0
+chem,0.514,0.368,0.137,0.152,0.551
+target,0.634,0.527,0.494,0.453,0.561
+transporter,0.605,0.484,0.427,0.358,0.558
+enzyme,0.615,0.497,0.459,0.431,0.537
+pathway,0.618,0.505,0.467,0.429,0.529
+indication,0.681,0.596,0.567,0.555,0.581
+sideeffect,0.604,0.551,0.41,0.418,0.606
+offsideeffect,0.66,0.647,0.51,0.502,0.657
diff --git a/.../processed/runs/replicatedIndividualScores_max_epochs-20-batch_size-200-H1-400-H2-300.csv b/.../processed/runs/replicatedIndividualScores_max_epochs-20-batch_size-200-H1-400-H2-300.csv
@@ -1,9 +1,9 @@
 Similarity,AUC,AUPR,F-measure,Recall,Precision
-chem,0.501,0.379,0.018,0.009,0.625
-target,0.731,0.72,0.621,0.572,0.707
-transporter,0.639,0.563,0.48,0.409,0.614
-enzyme,0.671,0.63,0.535,0.487,0.643
-pathway,0.706,0.691,0.587,0.56,0.66
-indication,0.781,0.788,0.692,0.672,0.74
-sideeffect,0.556,0.549,0.218,0.253,0.555
-offsideeffect,0.661,0.678,0.469,0.443,0.699
+chem,0.512,0.385,0.111,0.094,0.426
+target,0.643,0.535,0.51,0.48,0.559
+transporter,0.594,0.474,0.4,0.328,0.556
+enzyme,0.614,0.491,0.452,0.418,0.537
+pathway,0.625,0.507,0.483,0.461,0.524
+indication,0.685,0.607,0.572,0.558,0.591
+sideeffect,0.618,0.581,0.443,0.436,0.612
+offsideeffect,0.638,0.655,0.465,0.46,0.669
diff --git a/ddi/dataset.py b/ddi/dataset.py
@@ -95,10 +95,14 @@ def preprocess_labels(interaction_fpath):
 def create_setvector_features(X, num_sim_types):
     """reshape concatenated features from every similarity type matrix into set of vectors per ddi example"""
     e = X[np.newaxis, :, :]
+    # print('e.shape', e.shape)
     f = np.transpose(e, axes=(0, 2, 1))
+    # print('f.shape', f.shape)
     splitter = 2*num_sim_types 
     g = np.concatenate(np.split(f, splitter, axis=1), axis=0)
+    # print('g.shape', g.shape)
     h = np.transpose(g, axes=(2,0, 1))
+    # print('h.shape', h.shape)
     return h
 
 def get_stratified_partitions(ddi_datatensor, num_folds=5, random_state=42):

diff --git a/ddi/utilities.py b/ddi/utilities.py
@@ -39,6 +39,7 @@ def get_performance_results(similarity_type, target_dir, num_folds, dsettype):
                                 'fold_{}'.format(fold_num))
 
         score_file = os.path.join(fold_dir, 'score_{}.pkl'.format(dsettype))
+        # print(score_file)
         if os.path.isfile(score_file):
             mscore = ReaderWriter.read_data(score_file)
             perf_dict[0]['fold{}'.format(fold_num)] = mscore.s_auc
@@ -63,8 +64,9 @@ def build_performance_dfs(similarity_types, target_dir, num_folds, dsettype):
     aupr_df = pd.DataFrame()
     f1_df = pd.DataFrame()
     target_dir = create_directory(target_dir)
+    print(target_dir)
     for sim_type in similarity_types:
-        s_auc , s_aupr, s_f1 = get_performance_results(sim_type, target_dir, num_folds, dsettype)
+        s_auc, s_aupr, s_f1 = get_performance_results(sim_type, target_dir, num_folds, dsettype)
         auc_df = pd.concat([auc_df, s_auc], sort=True)
         aupr_df = pd.concat([aupr_df, s_aupr], sort=True)
         f1_df = pd.concat([f1_df, s_f1], sort=True)