Relationships between ligand binding and the shapes of the binding sites in families of homologous enzymes are investigated by comparing matrices of distances between key binding site atoms. Multiple linear regression is used to help identify key distances that influence ligand binding affinity. In order to illustrate the utility of this generic approach, we study protein kinase binding sites for ATP and the promiscuous competitive inhibitor, staurosporine. We show that the size of the gatekeeper residue and the closure between the first glycine of the GXGXXG motif and the aspartate of the DFG loop act together to promote tight binding. Our web-based tool, 'mapping analogous hetero-atoms onto residue interactions' (MAHORI), indicates that the greater the number of hydrogen bonds made by the kinase around the methylamine group of staurosporine, the tighter the binding. The conservation of surrounding atoms identified using our novel grid-based method clearly demonstrates that the most structurally conserved part of the binding site for staurosporine is the main chain of the hinge region. The critical role of interactions that are not dependent on side-chain identities is consistent with the promiscuous nature of this inhibitor.