Based on the similarity of the sequences of combinatorially selected peptides that have similar binding characteristics, we developed a bioinformatics approach that provides a general and simple methodology to quantitatively categorize a large number of inorganic binding peptides. The approach also provides a way to knowledge-based design a new set of binding sequences specific to inorganic surfaces with predictable functionalities. De novo designed peptides can then be expressed using genetic tools, such as redisplay, to assess the efficacy of the design via binding affinity evaluation. This process is analogous to the evolution process where successive cycles of mutation and selection lead to a progeny with improved functionality. Through the generation of new scoring matrices, our approach has the potential of constructing super-sequences specific to a group of noble metals, metal-oxides, or semiconductors as well as master sequences specific to individual substrates (e.g., a metal or an oxides). The binding characteristics of these de novo designed peptides could be further tailored via recombinant DNA technologies using, e.g., only the binding domains or multiple repeats of peptides, to create addressable molecular recognition, a great utility in nano- and bio-nanotechnology.