In this paper, we report a simple structure-based iterative optimizations (SUBITO) strategy to identify and optimize new protein ligands and inhibitors. The approach is based on a combination of NMR-based screening and computational docking methods and enabled the identification of novel chemical leads among hundreds of thousands of commercially available compounds by screening only a few hundred compounds from a scaffold library followed by iterative screening steps where only few dozen compounds are tested. As an application, we report on the discovery of a novel class of non-peptide reversible caspase inhibitors, with IC(50) values in the low micromolar range.