Successful design of hyperthermophilic proteins has given hope of rationally engineering the shapes and functions of novel polypeptide sequences. Nevertheless, formidable technical challenges remain. First, existing design methods require prior knowledge of a naturally-occurring, fixed-backbone conformation. An enormous gap exists between fixed-backbone approaches and true de novo design. Second, current design methods use empirical potential energy functions that cannot describe the physical chemistry of protein structure and enzymatic reactions. Finally, a conceptual basis for engineering specificity into molecular interactions is lacking. We are addressing these limitations within the context of the simplest of all protein folds, the coiled coil. We use parametric backbone models to generate and test novel protein backbone structures. We have developed and are characterizing a physically-based molecular mechanics potential that treats solvent as a dielectric continuum. The analytical form of the electrostatic potential is evaluated rapidly, allowing us to model the electrostatic effects of dynamic protein side-chain rearrangements. Finally, we are developing a general framework for sequence selection that specifies a sequence for a target structure by taking into consideration a large ensemble of competing conformational states. If successful, this approach will be applicable to almost any protein engineering problem.