Modeling of human corticosteroid binding globulin. Use of structure-activity relations in soft steroid binding to refine the structure.

PURPOSE: We propose a model for human corticosteroid binding globulin that is capable of explaining at the molecular level the experimentally observed binding affinities of four ligands. A new method of analyzing data from docking studies is proposed.
METHODS: Displacement of radioactive ligand by competitive binding gives the experimentally determined binding affinities of the competitors. A theoretical model, based on homology with crystallographically determined structures, was studied in an automated docking procedure for the determination of theoretical affinities. The docking runs were analyzed by a hybrid principal component-clustering analysis.
RESULTS: Of the two binding sites considered, only one--that in the vicinity of Cys 60--can reproduce the experimentally observed order of binding affinities although the lowest energies are found at the site in the vicinity of Cys 228.
CONCLUSIONS: Models proposed for proteins should be always conditioned to take into account experimentally observed results. In the current work, we have shown that an informed analysis of theoretical docking studies can lead to a more logical model of the protein, one that can explain and give a deeper understanding of the stereochemical requirements of binding.