add 'classification' column

nonadditivity/classification/classify.py

@@ -252,6 +252,37 @@ def _classify_circles(
     return classification
 
 
+def _is_surprisingly_nonadditive(circle: Circle) -> bool:
+    """Check whether a circle's nonadditivity is surprising.
+
+    Use keyword "circle" for this to work.
+
+    A circle is considered surprising i fnone of the following is true:
+    Distance between R groups ≤ 2 atoms
+    Tanimoto similarity of the transformation < 0.4
+    Number of exchanged heavy atoms > 10
+    Linker exchange transformations
+    Transformations with unassigned or inverted stereocenters
+
+    Args:
+        circle (Circle): circle to check.
+
+    Returns:
+        bool: True if surprising false if Mundane.
+    """
+    if circle.get_property(Circle.Properties.DISTANCE_BETWEEN_R_GROUPS) <= 2:
+        return False
+    if circle.get_property(Circle.Properties.MIN_TANIMOTO) < 0.4:
+        return False
+    if circle.get_property(Circle.Properties.NUM_HEAVY_ATOMS_DIFF) > 10:
+        return False
+    if circle.get_property(Circle.Properties.MAX_NUM_MMPDB_CUTS) > 1:
+        return False
+    return (
+        circle.get_property(Circle.Properties.HAS_INVERSION_IN_TRANSFORMATION) == "None"
+    )
+
+
 def _update_na_dataframe(
     na_dataframe: pd.DataFrame,
     circles: list[Circle],
@@ -294,6 +325,9 @@ def _update_na_dataframe(
         circles=circles,
         property_key=props.DISTANCE_BETWEEN_R_GROUPS,
     )
+    na_dataframe["classification"] = [
+        "surprising" if _is_surprisingly_nonadditive(c) else "mundane" for c in circles
+    ]
 
 
 def classify_circles(

nonadditivity/workflow/canonicalize_output.py

@@ -91,11 +91,15 @@ def _get_props() -> list[str]:
     Returns:
         list[str]: list of descripytion keys
     """
-    return ["num_atoms_between_r_groups", "ortho_classification"] + [
-        val
-        for val in Circle.classification_keys.values()
-        if "ortho" not in val and "between" not in val
-    ]
+    return (
+        ["num_atoms_between_r_groups", "ortho_classification"]
+        + [
+            val
+            for val in Circle.classification_keys.values()
+            if "ortho" not in val and "between" not in val
+        ]
+        + ["classification"]
+    )
 
 
 def _add_to_dict(
@@ -215,7 +219,6 @@ def canonicalize_na_dataframe(
             columns=[
                 "has_transformation_at_ortho",
                 "has_ortho_substituent_introduced",
-                "bonds_atoms_between_r_groups",
             ],
         )
     except KeyError:

tests/_test_files/test_naa_classify.csv

@@ -1,3 +1,3 @@
-Compound1,Compound2,Compound3,Compound4,SMILES1,SMILES2,SMILES3,SMILES4,Prop_Cpd1,Prop_Cpd2,Prop_Cpd3,Prop_Cpd4,Transformation1,Transformation2,Property,Series,Nonadditivity,Theo_Quantile,Circle_ID,hbond_donor_diff,hbond_acceptor_diff,formal_charge_diff,tpsa_diff,num_rot_bonds_diff,sp3_carbon_diff,log_p_diff,chi0_diff,chi1_diff,chi2_diff,num_heavy_atoms_diff,tertiary_amide_formed,has_stereoinversion_in_transformation,max_num_mmpdb_cuts,max_num_heavy_atom_in_transformation,compound_stereocenter_classification,min_transformation_tanimoto,substituents_in_same_ring_system,ortho_classification,num_atoms_between_r_groups
-ID1,ID2,ID4,ID3,CC=Cc1ccc(C)nc1,CC(F)=Cc1ccc(C)nc1,CC(F)=C(Cl)c1ccc(C)nc1,CC=C(Cl)c1ccc(C)nc1,4.5,4.2,5.9,3.7,[*:1][H]>>[*:1]F,[*:1][H]>>[*:1]Cl,TEST_PCHEMBL_VALUE,,2.5,0.5449521356173604,ID1_ID2_ID4_ID3_TEST_PCHEMBL_VALUE,0,0,0,0.0,0,0,0.8637000000000006,1.7404869760061565,0.7833624025851336,0.30279739252301363,2,False,None,1,1,None,0.4996873045653534,False,None,2
-ID1,ID4,ID6,ID5,CC=Cc1ccc(C)nc1,CC(F)=C(Cl)c1ccc(C)nc1,CC(F)=C(Cl)c1cnc(C)c(O)c1,CC=Cc1cnc(C)c(O)c1,4.5,5.9,4.0,3.4,[*:1]C=CC>>[*:1]C(Cl)=C(C)F,[*:1][H]>>[*:1]O,TEST_PCHEMBL_VALUE,,-0.8000000000000003,-0.5449521356173604,ID1_ID4_ID6_ID5_TEST_PCHEMBL_VALUE,1,1,0,20.230000000000004,0,0,1.158100000000001,2.610730464009235,1.1940460051080928,0.44307085693965886,3,False,None,1,5,None,0.40878552971576226,True,None,3
+Compound1,Compound2,Compound3,Compound4,SMILES1,SMILES2,SMILES3,SMILES4,Prop_Cpd1,Prop_Cpd2,Prop_Cpd3,Prop_Cpd4,Transformation1,Transformation2,Property,Series,Nonadditivity,Theo_Quantile,Circle_ID,hbond_donor_diff,hbond_acceptor_diff,formal_charge_diff,tpsa_diff,num_rot_bonds_diff,sp3_carbon_diff,log_p_diff,chi0_diff,chi1_diff,chi2_diff,num_heavy_atoms_diff,tertiary_amide_formed,has_stereoinversion_in_transformation,max_num_mmpdb_cuts,max_num_heavy_atom_in_transformation,compound_stereocenter_classification,min_transformation_tanimoto,substituents_in_same_ring_system,ortho_classification,num_atoms_between_r_groups,classification
+ID1,ID2,ID4,ID3,CC=Cc1ccc(C)nc1,CC(F)=Cc1ccc(C)nc1,CC(F)=C(Cl)c1ccc(C)nc1,CC=C(Cl)c1ccc(C)nc1,4.5,4.2,5.9,3.7,[*:1][H]>>[*:1]F,[*:1][H]>>[*:1]Cl,TEST_PCHEMBL_VALUE,,2.5,0.5449521356173604,ID1_ID2_ID4_ID3_TEST_PCHEMBL_VALUE,0,0,0,0.0,0,0,0.8637000000000006,1.7404869760061565,0.7833624025851336,0.30279739252301363,2,False,None,1,1,None,0.4996873045653534,False,None,2,mundane
+ID1,ID4,ID6,ID5,CC=Cc1ccc(C)nc1,CC(F)=C(Cl)c1ccc(C)nc1,CC(F)=C(Cl)c1cnc(C)c(O)c1,CC=Cc1cnc(C)c(O)c1,4.5,5.9,4.0,3.4,[*:1]C=CC>>[*:1]C(Cl)=C(C)F,[*:1][H]>>[*:1]O,TEST_PCHEMBL_VALUE,,-0.8000000000000003,-0.5449521356173604,ID1_ID4_ID6_ID5_TEST_PCHEMBL_VALUE,1,1,0,20.230000000000004,0,0,1.158100000000001,2.610730464009235,1.1940460051080928,0.44307085693965886,3,False,None,1,5,None,0.40878552971576226,True,None,3,surprising

tests/nonadditivity/classification/test_classify.py

@@ -12,6 +12,7 @@
     _create_compound_dict,
     _get_na_compounds,
     _get_transformations_for_na_dataframe,
+    _is_surprisingly_nonadditive,
     _update_na_dataframe,
     _update_per_compound_dataframe,
     classify,
@@ -37,7 +38,7 @@
     (Props.HAS_INVERSION_IN_TRANSFORMATION, "None"),
     (Props.MAX_NUM_MMPDB_CUTS, 1),
     (Props.MAX_HEAVY_ATOM_IN_TRANSFORMATION, 4),
-    (Props.COMPOUND_STEREO_CLASSIFICATION, "Unassigned"),
+    (Props.COMPOUND_STEREO_CLASSIFICATION, "Assigned"),
     (Props.MIN_TANIMOTO, 0.6626506024096386),
     (Props.SUBSTITUENT_ON_SAME_RING_SYSYTEM, False),
 ]
@@ -279,6 +280,33 @@ def test_classify_circle_special(
     assert result == [solution]
 
 
+@pytest.mark.parametrize(
+    "fixture_name, solution",
+    [
+        ("circle", True),
+        ("circle_2", True),
+        ("ortho_none_circle", False),
+        ("ortho_exchanged_circle", True),
+        ("ortho_both_circle", False),
+    ],
+)
+def test_is_surprisingly_nonadditive(
+    fixture_name: str,
+    solution: str,
+    request: pytest.FixtureRequest,
+) -> None:
+    """Test nonadditivity.classification.classify:_is_surprisingly_nonadditive.
+
+    Args:
+        fixture_name (str): fixture name
+        solution (str): expected output
+        request (pytest.FixtureRequest): pytest magic
+    """
+    circle: Circle = request.getfixturevalue(fixture_name)
+    circle.classify()
+    assert _is_surprisingly_nonadditive(circle) == solution
+
+
 @pytest.mark.parametrize(
     "prop, solution",
     [

tests/nonadditivity/workflow/test_canonicalize_output.py

@@ -84,7 +84,12 @@ def test_get_props() -> None:
         "num_atoms_between_r_groups",
     ):
         solution.remove(rem)
-    solution = ["num_atoms_between_r_groups", "ortho_classification", *solution]
+    solution = [
+        "num_atoms_between_r_groups",
+        "ortho_classification",
+        *solution,
+        "classification",
+    ]
     assert _get_props() == solution