We report here qualitative structure-activity relationship (SAR) studies for the molecular set called Tripos or Cramer steroid data set. These compounds are known to bind to corticosteroid binding globulin (CBG). In the present work we have used the electronic indices methodology (EIM). The EIM is based on Boolean relational rules exploring the concepts of local density of states and critical values for energy separation involving frontier orbitals. We have also carried out comparative principal component analysis (PCA) and hierarchical clustering analysis (HCA) studies with molecular descriptors obtained from EIM calculations. EIM, PCA and HCA correctly predict (100% accuracy) the steroid's biological activity. The present studies reinforce the universal applicability of the EIM descriptors and show that the combined use of EIM coupled to PCA can be a new efficient and powerful SAR tool. (C) 2002 Elsevier Science B.V. All rights reserved.619195205