A combination of the empirical valence bond method and a free energy perturbation approach is used to simulate the activity of genetically modified enzymes. The simulations reproduce in a semiquantitative way the observed effects of mutations on the activity and binding free energies of trypsin and subtilisin. This suggests that we are approaching a stage of quantitative structure-function correlation of enzymes. The analysis of the calculations points towards the electrostatic energy of the reacting system as the key factor in enzyme catalysis. The changes in the charges of the reacting system and the corresponding changes in "solvation" free energy (generalized here as the interaction between the charges and the given microenvironment) are emphasized. It is argued that a reliable evaluation of these changes might be sufficient for correlating structure and catalysis. The use of free energy perturbation methods and thermodynamic cycles for evaluation of solvation energies and reactivity is discussed, pointing out our early contributions. The apparent elaborated nature of our treatment is clarified, explaining that such a treatment is essential for consistent calculations of chemical reactions in polar environments. The problems associated with seemingly more rigorous quantum mechanical methods are discussed, emphasizing the inconsistency associated with using gas phase charge distributions. The importance of dynamic aspects is examined by evaluating the autocorrelation of the protein "reaction field" on the reacting substrate. It is found that, at least in the present case, dynamic effects are not important. The nature of the catalytic free energy is considered, arguing that the protein provides preoriented dipoles (polarized to stabilize the transition state charge distribution) and small reorganization energy, thus reducing the activation free energy. The corresponding catalytic free energy is related to the folding free energy, which is being invested in aligning the active site dipoles.