Generating and analyzing a dataset using RDKit

This is an exploratory notebook using rdkit and other libraries to examine melting points of hydrocarbons.

import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt

from rdkit import Chem
from rdkit.Chem import Descriptors, PandasTools

df = pd.read_csv("data/hydrocarbons.csv")
df.head()

	Class of hydrocarbon	IUPAC name	Melting point	Boiling point	Density@20å¡C*	Flash point	Autoignition temp	pubchem_id	smiles
0	Trimetylalkane	2,2,4-Trimethylpentane	-107.0	99.0	0.69	NaN	396	10907	CC(C)CC(C)(C)C
1	Triaromatics	Phenanthrene	99.0	338.0	1.18	171	>450	995	C1=CC=C2C(=C1)C=CC3=CC=CC=C32
2	Triaromatics	Anthracene	216.0	341.0	1.2825	NaN	NaN	8418	C1=CC=C2C=C3C=CC=CC3=CC2=C1
3	Triaromatics	1-methylanthracene	86.0	342.0	1.04799	NaN	NaN	11884	CC1=CC=CC2=CC3=CC=CC=C3C=C12
4	Triaromatics	2-methylanthracene	209.0	340.0	1.8	NaN	NaN	11936	CC1=CC2=CC3=CC=CC=C3C=C2C=C1

# Use pandas tools to 
PandasTools.AddMoleculeColumnToFrame(df,'smiles', 'Molecule', includeFingerprints=True)

df.head()

	Class of hydrocarbon	IUPAC name	Melting point	Boiling point	Density@20å¡C*	Flash point	Autoignition temp	pubchem_id	smiles	Molecule
0	Trimetylalkane	2,2,4-Trimethylpentane	-107.0	99.0	0.69	NaN	396	10907	CC(C)CC(C)(C)C	<rdkit.Chem.rdchem.Mol object at 0x7fc5f994c970>
1	Triaromatics	Phenanthrene	99.0	338.0	1.18	171	>450	995	C1=CC=C2C(=C1)C=CC3=CC=CC=C32	<rdkit.Chem.rdchem.Mol object at 0x7fc5f994ca50>
2	Triaromatics	Anthracene	216.0	341.0	1.2825	NaN	NaN	8418	C1=CC=C2C=C3C=CC=CC3=CC2=C1	<rdkit.Chem.rdchem.Mol object at 0x7fc5f994cba0>
3	Triaromatics	1-methylanthracene	86.0	342.0	1.04799	NaN	NaN	11884	CC1=CC=CC2=CC3=CC=CC=C3C=C12	<rdkit.Chem.rdchem.Mol object at 0x7fc5f994ccf0>
4	Triaromatics	2-methylanthracene	209.0	340.0	1.8	NaN	NaN	11936	CC1=CC2=CC3=CC=CC=C3C=C2C=C1	<rdkit.Chem.rdchem.Mol object at 0x7fc5f994cc80>

# Retrieve some descriptors

descriptors_df = pd.DataFrame()
descriptors_df["Melting point"] = df["Melting point"].copy()

descriptors_df["molecular_weight"] = df["Molecule"].apply(Chem.Descriptors.MolWt)
descriptors_df["num_aromatic_rings"] = df["Molecule"].apply(Chem.Descriptors.NumAromaticRings)

descriptors_df.head()

	Melting point	molecular_weight	num_aromatic_rings
0	-107.0	114.232	0
1	99.0	178.234	3
2	216.0	178.234	3
3	86.0	192.261	3
4	209.0	192.261	3

descriptors_melt = descriptors_df.melt(id_vars="Melting point")

descriptors_melt.head()

	Melting point	variable	value
0	-107.0	molecular_weight	114.232
1	99.0	molecular_weight	178.234
2	216.0	molecular_weight	178.234
3	86.0	molecular_weight	192.261
4	209.0	molecular_weight	192.261

g = sns.lmplot(data=descriptors_melt, y="Melting point", x="value", col="variable", col_wrap=2, facet_kws={"sharex": False})

Linear and Multilinear Regression Using SciKitLearn

Use this section to perform linear or multilinear regression using scikitlearn.

Fitting with Random Forest

from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import train_test_split

analyze = df.dropna(axis=0, subset="Melting point")

X = analyze[["molecular_weight", "num_aromatic_rings"]].to_numpy()
Y = analyze["Melting point"].to_numpy()

X_train, X_test, Y_train, Y_test = train_test_split(X, Y)

random_forest_model = RandomForestRegressor().fit(X_train, Y_train)
yerr = random_forest_model.score(X_test, Y_test)
print(yerr)

y_pred = random_forest_model.predict(X)

plt.figure()
plt.scatter(Y, y_pred)