Additional Features

ChemPlot offers additional features for chemical space visualization which can improve the understanding of the underlying similarities between the investigated molecules.

Using the Plotter object it is possible to create two different kind plots of the chemical space, aside from the scatterplots showed in the previous sections. These plots investigate the density distribution of the chemical space an are hexagonal bin plot and the kernel density estimate plot.

To show the before mentioned features we will use the BBBP (blood-brain barrier penetration) dataset 1, already mentioned in the previous section:

from chemplot import Plotter, load_data

data_BBBP = load_data("BBBP")
cp_BBBP = Plotter.from_smiles(data_BBBP["smiles"], target=data_BBBP["target"], target_type="C")

Hexagonal Bin Plot

In a hexagonal bin plot points are binned into hexagons, which in turn are coloured depending on the count of observations they cover. To create a hexagonal bin plot we need to pass the keyword “hex” as the kind parameter when visualizing the plot.

cp_BBBP.tsne(random_state=0)
cp_BBBP.visualize_plot(kind="hex")
../_images/tsne_hex.png

Kernel Density Estimate Plot

In a kernel density estimate plot, the data distribution is visualized by a continuous probability density curve which in our case is in 2 dimensions. To create a kernel density estimate plot we need to pass the keyword “kde” as the kind parameter when visualizing the plot.

cp_BBBP.visualize_plot(kind="kde")
../_images/tsne_kde.png

References:

1

Martins, Ines Filipa, et al. (2012). A Bayesian approach to in silico blood-brain barrier penetration modeling. Journal of chemical information and modeling 52.6, 1686-1697