The IA(3) polypeptide inhibitor from Saccharomyces cerevisiae interacts potently and selectively with its target, the S. cerevisiae vacuolar aspartic proteinase (ScPr). Upon encountering the enzyme, residues 2-32 of the intrinsically unstructured IA(3) polypeptide become ordered into an almost-perfect alpha-helix. In previous IA(3) mutagenesis studies, we identified important characteristics of the enzyme inhibitor interactions and generated a large dataset of variants with K(i) values determined experimentally at pH 3.1 and 4.7. Using this information, the three-dimensional structure of each variant was modelled in silico with the correct protonation for each experimental pH value. A set of descriptors of the inhibitor/ScPr interactions was then calculated and used to establish mathematical models relating the variant sequences to their inhibitory activities at each pH. Cross-validation, external-set validation and five separate selections of the training and test samples confirmed the robustness of the equations. A major contributor to the structure-activity relationship was the free energy of binding calculated by the FoldX program. The mathematical models were challenged further (i) by in silico alanine-scanning mutagenesis of residues 2-32 in IA(3) and relating binding energy to experimentally derived inhibition constants for selected representatives of these variants; and (ii) by predicting inhibitory-potencies for two novel IA(3)-variants. The predictions of the equations for these new IA(3)-variants with ScPr matched almost precisely the kinetic data determined experimentally. The models described represent valuable tools for the future design of novel inhibitor variants active against ScPr and other aspartic proteinases.