Table 2.

The data “cleaning” preprocess workflow.

Step	Action
Organic filter	Inorganic molecules are removed.
Element filter	Molecules containing elements other than C, H, O, N, P, S, Cl, Br, I, and F are removed.
Connectivity	Fragments except for the biggest from unconnected molecules are removed.
Standardizer	Directly bonded zwitterions are converted to the neutral representation. The charges on a molecule are set to a standard form.
Aromatizer	The molecules are converted into an aromatic form.
Canonicalize	The molecules are represented in canonical SMILES format.
Duplicate filter	Duplicate molecules are removed.