The inverse of a folding problem is to find the ideal sequence that folds into a particular protein structure. This problem has been addressed using the topology fingerprint-based threading algorithm, capable of calculating a score (energy) of an arbitrary sequence-structure pair. At first, the search is conducted by unconstrained minimization of the energy in sequence space. It is shown that using energy as the only design criterion leads to spurious solutions with incorrect amino acid composition. The problem lies in the general features of the protein energy surface as a function of both structure and sequence. The proposed solution is to design the sequence by maximizing the difference between its energy in the desired structure and in other known protein structures. Depending on the size of the database of structures 'to avoid', sequences bearing significant similarity to the native sequence of the target protein are obtained using this procedure.