An additivity-based sequence to reactivity algorithm for the
interaction of members of the Kazal family of protein inhibitors with
six selected serine proteinases is described. Ten consensus variable
contact positions in the inhibitor were identified, and the 19 possible
variants at each of these positions were expressed. The free energies
of interaction of these variants and the wild type were measured. For
an additive system, this data set allows for the calculation of all
possible sequences, subject to some restrictions. The algorithm was
extensively tested. It is exceptionally fast so that all possible
sequences can be predicted. The strongest, the most specific possible,
and the least specific inhibitors were designed, and an evolutionary
problem was solved.