Pipelining¶

scikit-chem expands on the scikit-learn Pipeline object to support filtering. It is initialized using a list of Transformer objects.

In [10]:

pipeline = skchem.pipeline.Pipeline([
        skchem.standardizers.ChemAxonStandardizer(keep_failed=True),
        skchem.forcefields.UFF(),
        skchem.filters.OrganicFilter(),
        skchem.descriptors.MorganFeaturizer()])

The pipeline will apply each in turn to objects, using the the highest priority function that each object implements, according to the order transform_filter > filter > transform.

For example, our pipeline can transform sodium acetate all the way to fingerprints:

In [11]:

mol = skchem.Mol.from_smiles('CC(=O)[O-].[Na+]')

In [4]:

pipeline.transform_filter(mol)

Out[4]:

morgan_fp_idx
     0
     0
     0
     0
     0
       ..
  0
  0
  0
  0
  0
Name: MorganFeaturizer, dtype: uint8

It also works on collections of molecules:

In [12]:

mols = skchem.read_smiles('https://archive.org/download/scikit-chem_example_files/example.smi', name_column=1); mols

Out[12]:

batch
ethane                          <Mol: CC>
propane                        <Mol: CCC>
benzene                   <Mol: c1ccccc1>
sodium acetate    <Mol: CC(=O)[O-].[Na+]>
serine                <Mol: NC(CO)C(=O)O>
Name: structure, dtype: object

In [16]:

pipeline.transform_filter(mols)

ChemAxonStandardizer: 100% (5 of 5) |##########################################| Elapsed Time: 0:00:04 Time: 0:00:04
UFF: 100% (5 of 5) |###########################################################| Elapsed Time: 0:00:00 Time: 0:00:00
OrganicFilter: 100% (5 of 5) |#################################################| Elapsed Time: 0:00:00 Time: 0:00:00
MorganFeaturizer: 100% (5 of 5) |##############################################| Elapsed Time: 0:00:00 Time: 0:00:00

Out[16]:

morgan_fp_idx	0	1	2	3	4	...	2043	2044	2045	2046	2047
batch
ethane	0	0	0	0	0	...	0	0	0	0	0
propane	0	0	0	0	0	...	0	0	0	0	0
benzene	0	0	0	0	0	...	0	0	0	0	0
sodium acetate	0	0	0	0	0	...	0	0	0	0	0
serine	0	0	0	0	0	...	0	0	1	0	0

5 rows × 2048 columns

In [ ]: