A computational technique has been developed for the construction of nb initio quality electron density distributions for large peptides and proteins. A database of a specialized set of molecular density fragments is constructed for use with the Molecular Electron Density Lego Assembler (MEDLA) method for building molecular electron densities of biopolymers composed from amino acids. The MEDLA program uses ab initio electron densities from the molecular fragment database and a set of atomic coordinates available, e.g., from X-ray diffraction experiments or from a molecular modeling program such as BIOGRAF, to construct the electron density for any specified conformation of the molecule. The current database can be used to compute the electron density distributions for any peptides and proteins that are made up of the 20 most common amino acids. The MEDLA program generates ab initio quality, three-dimensional electron densities for much larger molecules than those which could be computed at present using conventional ab initio methods. Even for molecules of > 1000 atoms, the MEDLA method requires minimal computational time. The method generates the electronic density for the entire molecule or if desired for any specific molecular fragment such as the backbone of a protein. The entire range of the electron density distribution is computed, from which molecular isodensity contour (MIDCO) surfaces for any density threshold value can be constructed using AVS or other visualization packages. The MIDCO surfaces provide a far more realistic description of molecular shape than the commonly used fused sphere Van der Waals surfaces or solvent accessible surfaces based on spherical atom models. In this work, nb initio quality electron densities are calculated for simple model peptides, for some important bioactive peptides in low-energy conformations, for the globular protein crambin comprised of 642 atoms in 46 amino acid residues, and for the gene 5 protein, made up of 87 residues with a total of 1384 atoms.