Other Splitters

• MolecularWeightSplit
• StratifiedDistributionSplit

In [1]:

%load_ext autoreload
%autoreload 2

import datamol as dm
import splito

from utils import visualize_chemspace

%load_ext autoreload %autoreload 2 import datamol as dm import splito from utils import visualize_chemspace

MolecularWeightSplit¶

Splits the dataset by sorting the molecules by their molecular weight and then finding an appropriate cutoff to split the molecules in two sets. The application use-case is to train a model which is able to learn the SAR from small molecules and aim to generalize the larger molecules.

In [9]:

# load dataset
data = dm.data.chembl_drugs()
data.head(5)

# load dataset data = dm.data.chembl_drugs() data.head(5)

Out[9]:

	smiles
0	Br.CC(N)Cc1ccc(O)cc1
1	Br.CC(NC(C)(C)C)C(=O)c1cccc(Cl)c1
2	Br.CC1C2Cc3ccc(O)cc3C1(C)CCN2CCc1ccccc1
3	Br.CCN(c1cc(-c2ccc(CN3CCOCC3)cc2)cc(C(=O)NCc2c...
4	Br.CN(C)CCCC1(c2ccc(F)cc2)OCc2cc(C#N)ccc21

In [10]:

splitter = splito.MolecularWeightSplit(test_size=0.2, random_state=111, generalize_to_larger=True)
train_idx, test_idx = next(splitter.split(X=data.smiles.tolist()))
assert train_idx.shape[0] > test_idx.shape[0]

data.loc[train_idx, "MolecularWeightSplit"] = "train"
data.loc[test_idx, "MolecularWeightSplit"] = "test"

splitter = splito.MolecularWeightSplit(test_size=0.2, random_state=111, generalize_to_larger=True) train_idx, test_idx = next(splitter.split(X=data.smiles.tolist())) assert train_idx.shape[0] > test_idx.shape[0] data.loc[train_idx, "MolecularWeightSplit"] = "train" data.loc[test_idx, "MolecularWeightSplit"] = "test"

2023-09-22 08:59:53.879 | WARNING  | splito._molecular_weight:_iter_indices:66 - n_splits=5 > 1, but MolecularWeightSplit is deterministic and will always return the same split!

In [11]:

data["mw"] = data.smiles.apply(lambda smi: dm.descriptors.mw(dm.to_mol(smi)))

data["mw"] = data.smiles.apply(lambda smi: dm.descriptors.mw(dm.to_mol(smi)))

In [12]:

data.groupby(by="MolecularWeightSplit")["mw"].hist(legend=True)

data.groupby(by="MolecularWeightSplit")["mw"].hist(legend=True)

Out[12]:

MolecularWeightSplit
test     Axes(0.125,0.11;0.775x0.77)
train    Axes(0.125,0.11;0.775x0.77)
Name: mw, dtype: object

In [13]:

visualize_chemspace(data, split_names=["MolecularWeightSplit"])

visualize_chemspace(data, split_names=["MolecularWeightSplit"])

Out[13]:

<Figure size 640x480 with 0 Axes>

<Figure size 640x480 with 0 Axes>

StratifiedDistributionSplit¶

Split a dataset using the values of a readout, so both train, test and valid have the same distribution of values. Instead of binning using some kind of interval, we will instead use a 1D clustering of the readout.

In [14]:

# load dataset
data = dm.data.freesolv()
data.head(5)

# load dataset data = dm.data.freesolv() data.head(5)

Out[14]:

	iupac	smiles	expt	calc
0	4-methoxy-N,N-dimethyl-benzamide	CN(C)C(=O)c1ccc(cc1)OC	-11.01	-9.625
1	methanesulfonyl chloride	CS(=O)(=O)Cl	-4.87	-6.219
2	3-methylbut-1-ene	CC(C)C=C	1.83	2.452
3	2-ethylpyrazine	CCc1cnccn1	-5.45	-5.809
4	heptan-1-ol	CCCCCCCO	-4.21	-2.917

In [15]:

splitter = splito.StratifiedDistributionSplit(test_size=0.2, random_state=111)
train_idx, test_idx = next(splitter.split(X=data["smiles"].tolist(), y=data.expt.tolist()))
assert train_idx.shape[0] > test_idx.shape[0]

data.loc[train_idx, "StratifiedDistributionSplit"] = "train"
data.loc[test_idx, "StratifiedDistributionSplit"] = "test"

splitter = splito.StratifiedDistributionSplit(test_size=0.2, random_state=111) train_idx, test_idx = next(splitter.split(X=data["smiles"].tolist(), y=data.expt.tolist())) assert train_idx.shape[0] > test_idx.shape[0] data.loc[train_idx, "StratifiedDistributionSplit"] = "train" data.loc[test_idx, "StratifiedDistributionSplit"] = "test"

Check the distributions of train/test set

In [16]:

data.groupby(by="StratifiedDistributionSplit")["expt"].plot.density(legend=True)

data.groupby(by="StratifiedDistributionSplit")["expt"].plot.density(legend=True)

Out[16]:

StratifiedDistributionSplit
test     Axes(0.125,0.11;0.775x0.77)
train    Axes(0.125,0.11;0.775x0.77)
Name: expt, dtype: object

In [17]:

visualize_chemspace(data, split_names=["StratifiedDistributionSplit"])

visualize_chemspace(data, split_names=["StratifiedDistributionSplit"])

Out[17]:

<Figure size 640x480 with 0 Axes>

<Figure size 640x480 with 0 Axes>

---

• The End :-)