The immediate result of any attempt of molecular modelling, either by automated procedures or by more manual methodes is a set of coordinates which can, by appropriate graphics programs, be translated into a set of pretty pictures to adorn a paper, theses or talk.

However, what is usually expected from a modelling project is more: A scientist asking for a model of his compound (or for an experimental structure determination, for that matter) hopes to get answers to specific questions about the physical properties of his protein: How can I get better production yield or higher thermodynamic stability. How can the affinity and/or the specificity of my antibody fragment be improved, or why did I loose affinity in a attempted humanization? Which of the dozens of mutations generated in a gene shuffling experiment are significant? What is the most efficent randomization strategy if I wish to improve a certain property?

Many of these questions cannot be answered from the structure alone. They depend on the properties of inaccessible folding intermediates, precise information of antibody-antigen interaction, factors affecting the antibody in an in-vivo environment.

To answer some of these questions, the model serves as a focussing point in a detailed analysis of all aspects of a given molecule, experimental results as well as theoretical calculations=> a few examples