DrugESP: A Quantum Chemical Dataset of Electrostatic and Electronic Properties of Drug-like Molecules

DrugESP: A Quantum Chemical Dataset of Electrostatic and Electronic Properties of Drug-like Molecules DrugESP-149K provides quantum-mechanically derived electronic and electrostatic properties for 149,074 drug-like small molecules sampled from the MOSES dataset, computed at the B3LYP/6-31G* level of theory (ORCA 6.0.0). Each molecule includes an MMFF94-optimized 3D geometry, CHELPG and Mulliken atomic partial charges, dipole/quadrupole moments, isotropic and tensor polarizability, HOMO/LUMO energies and gap, and electrostatic potential extrema (Vmin/Vmax) on the COSMO van der Waals surface. The release additionally includes three extensions built on the same core methodology: - A charged-species set (20,401 molecules) covering protonation and deprotonation states of a sample of the core dataset's parent molecules, generated via pKa-aware protonation-state enumeration (dimorphite-dl) at physiological pH and computed at B3LYP/6-31+G*/CPCM(water) to account for implicit solvation and diffuse electron density.- A matched neutral reference set (1,780 molecules), recomputing the corresponding neutral parents at the same B3LYP/6-31+G*/CPCM(water) level, enabling direct comparison across charge states without confounding level-of-theory differences.- An FDA-approved reference set (1,247 neutral molecules) sourced from DrugCentral and independently cross-verified against the FDA Orange Book, each entry carrying its regulatory application number, type, and trade name. A validation supplement is also included: 500 randomly sampled core-dataset molecules with their raw ORCA electrostatic potential grids (.vpot files), supporting independent reproduction of the CHELPG-versus-ab-initio ESP validation reported for this dataset. All data are provided as JSON, with a shared schema documented in the accompanying metadata file.