Quantitative structure-activity relationships by distance geometry: systematic analysis of dihydrofolate reductase inhibitors.

Extensions are presented for the distance geometry approach to rationalizing ligand binding data. These are algorithms to (i) detect when homologues are not binding with the same orientation in the binding site although they are chemically similar; (ii) deduce what the binding site's size and shape must be; and (iii) calculate the optimal set of interaction energies between parts of the site and parts of the ligand molecules. This improved methodology is tested on a set of 68 quinazoline inhibitors of S. faecium dihydrofolate reductase. Results are discussed and compared with the Hansch method of QSAR, and an improved inhibitor is predicted.