The current work outlines modeling strategies used to support the discovery of potent caspase-3 inhibitors. First, a group of caspase DEVD aldehyde crystal structure complexes were aligned. Caspases-1, 3, 7, and 8 were considered. Residues proximal to DEVD aldehyde define the binding pockets of the various caspases. It was found that SER 343 and PHE 381B in the S3 and S4 pockets were exclusive to caspase-3. Therefore targeting ligand interactions with these residues provides a general recipe for making caspase-3 selective ligands. Selective or pan inhibition, however, is better understood by examination of the detailed interactions that ligands make with the caspase proteins. To show how this is done in a qualitatively fashion, two well-known ligands were modeled into the caspase-3 and caspase-7 binding pockets. These molecules were M13 (potent at Caspase-3, not at caspase-7) and M826 (potent at both caspase-3 and caspase-7). A visual inspection of the modeled geometries suggests that in the case of M13 the key residue for caspase-3 selectivity over caspse-7 is ASN_342. In the case of M826 the key residue for potency at both caspase-3 and caspase-7 is ARG_341. For an ultimate understanding of the ligand-protein binding one must map some kind of binding thermodynamics onto modeled geometries. Therefore, a methodology was derived for the calculation of ligand-caspase binding free energies associated with modeled geometries. The methodology was validated by showing that calculated binding free energy differences correlated reasonably well (R2 = 0.87) with experimentally determined Ki ratios for a set of pyrimidoindolone (3,4-dihydropyrimido (1,2-a) indol-10 (2H)-one) caspase-3 inhibitors. We believe this broad approach (identification of target residues, visual examination of ligand-protein interactions, and binding free energy calculation) can be useful for the effective design of selective or pan caspase inhibitors.