Skip to content

taps.apps.moldesign.chemfunctions

taps/apps/moldesign/chemfunctions.py

MorganFingerprintTransformer 

MorganFingerprintTransformer(
    length: int = 256, radius: int = 4
)

Bases: BaseEstimator, TransformerMixin

Class that converts SMILES strings to fingerprint vectors.

Source code in taps/apps/moldesign/chemfunctions.py 
171
172
173
def __init__(self, length: int = 256, radius: int = 4) -> None:
    self.length = length
    self.radius = radius

fit() 

fit(X: Any, y: Any = None) -> Self

Fit the transformer.

Source code in taps/apps/moldesign/chemfunctions.py 
175
176
177
def fit(self, X: Any, y: Any = None) -> Self:  # noqa: N803
    """Fit the transformer."""
    return self  # Do need to do anything

transform() 

transform(X: Any, y: Any = None) -> Any

Compute the fingerprints.

Parameters:

  • X (Any) –

    List of SMILES strings.

  • y (Any, default: None ) –

    Ignored.

Returns:

  • Any –

    Array of fingerprints.

Source code in taps/apps/moldesign/chemfunctions.py 
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
def transform(self, X: Any, y: Any = None) -> Any:  # noqa: N803
    """Compute the fingerprints.

    Args:
        X: List of SMILES strings.
        y: Ignored.

    Returns:
        Array of fingerprints.
    """
    my_func = partial(
        compute_morgan_fingerprints,
        fingerprint_length=self.length,
        fingerprint_radius=self.radius,
    )
    with ProcessPoolExecutor(max_workers=n_workers) as pool:
        fing = list(pool.map(my_func, X, chunksize=2048))
    return np.vstack(fing)

generate_initial_xyz() 

generate_initial_xyz(mol_string: str) -> str

Generate the XYZ coordinates for a molecule.

Parameters:

  • mol_string (str) –

    SMILES string.

Returns:

  • str –

    XYZ coordinates for the molecule.

Source code in taps/apps/moldesign/chemfunctions.py 
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
def generate_initial_xyz(mol_string: str) -> str:
    """Generate the XYZ coordinates for a molecule.

    Args:
        mol_string: SMILES string.

    Returns:
        XYZ coordinates for the molecule.
    """
    # Generate 3D coordinates for the molecule
    mol = Chem.MolFromSmiles(mol_string)
    if mol is None:
        raise ValueError(f'Parse failure for {mol_string}')
    mol = Chem.AddHs(mol)
    AllChem.EmbedMolecule(mol, randomSeed=1)
    AllChem.MMFFOptimizeMolecule(mol)

    # Save geometry as 3D coordinates
    xyz = f'{mol.GetNumAtoms()}\n'
    xyz += mol_string + '\n'
    conf = mol.GetConformer()
    for i, a in enumerate(mol.GetAtoms()):
        s = a.GetSymbol()
        c = conf.GetAtomPosition(i)
        xyz += f'{s} {c[0]} {c[1]} {c[2]}\n'

    return xyz

compute_vertical() 

compute_vertical(smiles: str) -> float

Run the ionization potential computation.

Parameters:

  • smiles (str) –

    SMILES string to evaluate.

Returns:

  • float –

    Ionization energy in Ha.

Source code in taps/apps/moldesign/chemfunctions.py 
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
def compute_vertical(smiles: str) -> float:
    """Run the ionization potential computation.

    Args:
        smiles: SMILES string to evaluate.

    Returns:
        Ionization energy in Ha.
    """
    # Make the initial geometry
    xyz = generate_initial_xyz(smiles)

    # Make the XTB calculator
    calc = XTB(accuracy=0.05)

    # Parse the molecule
    atoms = read(StringIO(xyz), format='xyz')

    # Compute the neutral geometry. Uses QCEngine
    # (https://github.com/MolSSI/QCEngine) to handle interfaces to XTB.
    atoms.calc = calc
    dyn = LBFGSLineSearch(atoms, logfile=None)
    dyn.run(fmax=0.02, steps=250)

    neutral_energy = atoms.get_potential_energy()

    # Compute the energy of the relaxed geometry in charged form
    charges = np.ones((len(atoms),)) * (1 / len(atoms))
    atoms.set_initial_charges(charges)
    charged_energy = atoms.get_potential_energy()

    return charged_energy - neutral_energy

compute_morgan_fingerprints() 

compute_morgan_fingerprints(
    smiles: str,
    fingerprint_length: int,
    fingerprint_radius: int,
) -> ndarray

Get Morgan Fingerprint of a specific SMILES string.

Adapted from: https://github.com/google-research/google-research/blob/dfac4178ccf521e8d6eae45f7b0a33a6a5b691ee/mol_dqn/chemgraph/dqn/deep_q_networks.py#L750

Parameters:

  • smiles (str) –

    The molecule as a SMILES string.

  • fingerprint_length (int) –

    Bit-length of fingerprint.

  • fingerprint_radius (int) –

    Radius used to compute fingerprint.

Returns:

  • ndarray –

    Array with shape [hparams, fingerprint_length] of the Morgan fingerprint.

Source code in taps/apps/moldesign/chemfunctions.py 
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
def compute_morgan_fingerprints(
    smiles: str,
    fingerprint_length: int,
    fingerprint_radius: int,
) -> np.ndarray:
    """Get Morgan Fingerprint of a specific SMILES string.

    Adapted from:
    https://github.com/google-research/google-research/blob/dfac4178ccf521e8d6eae45f7b0a33a6a5b691ee/mol_dqn/chemgraph/dqn/deep_q_networks.py#L750

    Args:
        smiles: The molecule as a SMILES string.
        fingerprint_length: Bit-length of fingerprint.
        fingerprint_radius: Radius used to compute fingerprint.

    Returns:
        Array with shape `[hparams, fingerprint_length]` of the Morgan \
        fingerprint.
    """
    # Parse the molecule
    molecule = Chem.MolFromSmiles(smiles)

    # Compute the fingerprint
    fingerprint = AllChem.GetMorganFingerprintAsBitVect(
        molecule,
        fingerprint_radius,
        fingerprint_length,
    )
    arr = np.zeros((1,), dtype=np.bool_)

    # ConvertToNumpyArray takes ~ 0.19 ms, while
    # np.asarray takes ~ 4.69 ms
    DataStructs.ConvertToNumpyArray(fingerprint, arr)
    return arr

train_model() 

train_model(
    smiles: list[str], properties: list[float]
) -> Pipeline

Train a machine learning model using Morgan Fingerprints.

Parameters:

  • smiles (list[str]) –

    SMILES strings for each molecule

  • properties (list[float]) –

    List of a property for each molecule

Returns:

  • Pipeline –

    A trained model.

Source code in taps/apps/moldesign/chemfunctions.py 
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
def train_model(smiles: list[str], properties: list[float]) -> Pipeline:
    """Train a machine learning model using Morgan Fingerprints.

    Args:
        smiles: SMILES strings for each molecule
        properties: List of a property for each molecule

    Returns:
        A trained model.
    """
    model = Pipeline(
        [
            ('fingerprint', MorganFingerprintTransformer()),
            (
                'knn',
                KNeighborsRegressor(
                    n_neighbors=4,
                    weights='distance',
                    metric='jaccard',
                    n_jobs=-1,
                ),
            ),
        ],
    )

    return model.fit(smiles, properties)

run_model() 

run_model(model: Any, smiles: list[str]) -> DataFrame

Run a model on a list of smiles strings.

Parameters:

  • model (Any) –

    Trained model that takes SMILES strings as inputs.

  • smiles (list[str]) –

    List of molecules to evaluate.

Returns:

  • DataFrame –

    A dataframe with the molecules and their predicted outputs.

Source code in taps/apps/moldesign/chemfunctions.py 
227
228
229
230
231
232
233
234
235
236
237
238
def run_model(model: Any, smiles: list[str]) -> pd.DataFrame:
    """Run a model on a list of smiles strings.

    Args:
        model: Trained model that takes SMILES strings as inputs.
        smiles: List of molecules to evaluate.

    Returns:
        A dataframe with the molecules and their predicted outputs.
    """
    pred_y = model.predict(smiles)
    return pd.DataFrame({'smiles': smiles, 'ie': pred_y})