Dimeric glutathione S-transferases (GSTs) are pharmacological targets for several diseases, including cancer. Isoform specificity has been difficult to achieve due to their overlapping substrate selectivity. Here we demonstrate the utility of bivalent GST inhibitors and their optimization via combinatorial linker design. A combinatorial library with dipeptide linkers emanating symmetrically from a central scaffold (bis-3,5-aminomethyl benzoic acid, AMAB) to connect two ethacrynic acid moieties was prepared and decoded via iterative deconvolution, against the isoforms GSTA1-1 and GSTP1-1. The library yielded high affinity GSTA1-1 selective inhibitors (70-120-fold selectivity) and with stoichiometry of one inhibitor: one GSTA1-1 dimer. Saturation Transfer Difference (STD) NMR with one of these inhibitors, with linker structure (Asp-Gly-AMAB-Gly-Asp) and K(D) = 42 nM for GSTA1-1, demonstrates that the Asp-Gly linker interacts tightly with GSTA1-1, but not P1-1. H/D exchange mass spectrometry was used to map the protein binding site and indicates that peptides within the intersubunit cleft and in the substrate binding site are protected by inhibitor from solvent exchange. A model is proposed for the binding orientation of the inhibitor, which is consistent with electrostatic complementarity between the protein cleft and inhibitor linker as the source of isoform selectivity and high affinity. The results demonstrate the utility of combinatorial, or "irrational", linker design for optimizing bivalent inhibitors.