make the test work without target data, even when normalization is se…

…t to True

deeprank/generate/NormalizeData.py

@@ -127,22 +127,25 @@ def _extract_data(self):
                     self.parameters['features'][feat_types][name].add(
                         np.mean(mat), np.var(mat))
 
-            # get the target groups
-            target_group = f5.get(mol + '/targets')
+            mol_group = f5[mol]
+            if "targets" in mol_group:
 
-            # loop over all the targets
-            for tname, tval in target_group.items():
+                # get the target groups
+                target_group = mol_group["targets"]
 
-                # we skip the already computed target
-                if tname in self.skip_target:
-                    continue
+                # loop over all the targets
+                for tname, tval in target_group.items():
 
-                # create a new item if needed
-                if tname not in self.parameters['targets']:
-                    self.parameters['targets'][tname] = MinMaxParam()
+                    # we skip the already computed target
+                    if tname in self.skip_target:
+                        continue
+
+                    # create a new item if needed
+                    if tname not in self.parameters['targets']:
+                        self.parameters['targets'][tname] = MinMaxParam()
 
-                # update the value
-                self.parameters['targets'][tname].update(tval[()])
+                    # update the value
+                    self.parameters['targets'][tname].update(tval[()])
 
         f5.close()
 

deeprank/learn/DataSet.py

@@ -761,7 +761,10 @@ def _read_norm(self):
                         mean, var)
 
             # handle the target
-            if self.select_target is not None:
+            if self.select_target is not None and \
+                    "targets" in data and \
+                    self.select_target in data["targets"]:
+
                 minv = data['targets'][self.select_target].min
                 maxv = data['targets'][self.select_target].max
                 self.param_norm['targets'][self.select_target].update(minv)

test/test_learn.py

@@ -31,19 +31,19 @@ def test_predict_without_target():
        'number_of_points': [30,30,30],
        'resolution': [1.,1.,1.],
        'atomic_densities': atomic_densities,
-    }   
+    }
 
     environment = Environment(pdb_root="test/data/pdb", pssm_root="test/data/pssm")
 
     variants = [PdbVariantSelection("101m", "A", 10, valine, cysteine,
                                     protein_accession="P02144", protein_residue_number=10,
-                                    variant_class=VariantClass.BENIGN),
+                                    ),
                 PdbVariantSelection("101m", "A", 8, glutamine, cysteine,
                                     protein_accession="P02144",
-                                    variant_class=VariantClass.PATHOGENIC),
+                                    ),
                 PdbVariantSelection("101m", "A", 9, glutamine, cysteine,
                                     protein_accession="P02144", protein_residue_number=9,
-                                    variant_class=VariantClass.PATHOGENIC)]
+                                    )]
     augmentation = 5
 
     work_dir_path = mkdtemp()
@@ -64,8 +64,7 @@ def test_predict_without_target():
 
         dataset = DataSet(hdf5_path, grid_info=grid_info,
                           select_feature='all',
-                          select_target='target1',
-                          normalize_features=False)
+                          normalize_features=True)
 
         eq_(len(dataset), len(variants) * (augmentation + 1))
         ok_(dataset[0] is not None)