A simple study of the information available within proteins of known structure reveals both the limits of structure prediction based on overall statistical correlations between sequence and structure, and the areas where there is still a possibility for further structural database exploitation.
