powered by
A dataset containing SMILES of 5,000 compounds from PubChem
trainedSMI
A character vector with 5,000 rows of 1 variable: