Reference Python implementation of Delaney's ESOL method — a multiple linear regression on four RDKit-computable descriptors (molecular weight, cLogP, number of rotatable bonds, aromatic proportion) that predicts aqueous solubility (log S, mol/L) directly from a SMILES string. Repository contains a refit of the original 1004-compound Delaney dataset using modern RDKit descriptors plus a DeepChem-based benchmark comparison.