The development of a theoretical databank of transferable pseudoatoms for fast prediction of the electron densities and related electronic properties of proteins is described. Chemically unique pseudoatoms identified on the basis of common connectivity and bonding are extracted from ab initio molecular densities of a large number of small molecules using a least-squares projection technique in Fourier transform space. The performance of the databank is evaluated by comparison of the electron densities and electrostatic properties of the amino acids GLN, SER, and LEU and their dimers with those obtained from molecular calculations on the same test compounds. It is found that deformation density bond peaks are reproduced to within 0.02-0.10 e/Angstrom(3), whereas electrostatic potentials, bond critical point indices, atomic charges, and molecular moments show differences with results from calculations performed directly on the test molecules which are comparable with or smaller than the spread of the values between different ab initio methods (Hamiltonian, basis set, etc.). The order of intermolecular electrostatic interaction energies for selected dimers of the test compounds are well reproduced, though the results are always smaller, by about 25 kJ/mol on average, than electrostatic energies from Morokuma-Ziegler decomposition of the total interaction energy evaluated with the ADF program. The difference is attributed to the limitations of the Buckingham-type approximation for electrostatic interactions, used in the current study, which assumes nonoverlapping charge densities. The consistency achieved by the pseudoatom databank is much better than that obtained with the AMBER99, CHARMM27, MM3, and MMFF94 force fields, which sometime overestimate, sometimes underestimate, the electrostatic interaction energy. The electrostatic component of the binding energies (directly related to the enthalpy of sublimation) of molecules in crystals, calculated based on the databank parameters, agree within 25-60 kJ/mol with the total binding energies evaluated ab initio at the Density Functional level of theory, even though the exchange-repulsion and dispersion terms have not been taken into account in the databank values.