Variable mapping of structure-activity relationships: Application to 17-spirolactone derivatives with mineralocorticoid activity

Fifty-four steroid homologs, belonging to the series of 17-spirolactones, were modelled by molecular and quantum mechanics. We studied the affinity of these compounds for the cytosolic mineralocorticoid receptor by way of various parameters describing each structure and its molecular properties. After the failure of a classic preliminary QSAR study, demonstrating the nonlinear relationships between affinity and structural descriptors, we constructed a model allowing us to predict the affinity of new compounds. Our method is based on simple graphic tools coupled to a cluster significance analysis. A complementary study of the activity relating the prediction of the antagonist/agonist character of 37 high-affinity compounds was also carried out using the same methodology. The principal electronic and structural characteristics leading to a selective activity were revealed.