A new method for calculating geometry-dependent atomic charges (GDACs) for polypeptides is presented. It overcomes the limitations of the partial equalization of orbital electronegativity (PEOE) and modified PEOE (MPEOE) methods, which depend only on connectivity, not geometry. Introduction of distance-dependent damping factors helps to include the effect of environment in determining the variation of bond distance (without explicit contribution of the correlated variation of bond angles), and thereby to reduce the number of parameters required to represent different atomic species. Since the correlation between the geometry and the dipole moments of molecules is crucial for this method, ab initio molecular orbital calculations were carried out to obtain the geometries and dipole moments with the 6-31G** basis set at the level of B3LYP theory. When bond distances are fixed prior to a charge calculation, the methodology outlined here leads to a direct calculation of the permanent molecular charge distribution represented as a set of distributed monopoles that depend on the geometry of the molecule. Hence, this method automatically accounts for the transferability of charges of small amino acid residues to build up a large polypeptide molecule, and can therefore provide an approximate description of any redistribution of charge density of large polypeptide molecules. The parameters characterizing the charge transfer in the formation of bonds were optimized by using dipole moment components and total dipole moments of 50 molecules that serve as models for the backbone and side chains of proteins. The calculated total dipole moments of these 50 molecules agree well with the ab initio results within an error of 5%. The new charge scheme has been applied to seven conformers of N ' -acetylalanine-N ' -methylamide (Ac-Ala-NHMe) with good agreement between ab initio and GDAC dipole moments. This method, however, gives poor results for conjugated systems that are larger than amides.