Ellipticine is a molecule derived from the natural extract Ochrosia elliptica. This molecule and its derivatives are highly cytotoxic to malignant cultured cells. The relatively simple structure of ellipticine has prompted chemists to design various structural modifications in order to obtain either more active derivatives or information on the structural moieties required for pharmacological activities. In the present work we report theoretical structure-activity relationship studies for 40 ellipticine derivatives using pattern-recognition methods such as electronics indices methodology (EIM), principal component analysis (PCA) and hierarchical clustering analysis (HCA) with molecular descriptors obtained from semiempirical parametric method 3 (PM3) calculations. By applying selected molecular descriptors it was possible to classify active and inactive compounds with accuracy up to 92% and also to suggest the activity of new untested molecules. These descriptors have been only recently discussed in the literature as new possible universal parameters for defining the biological activity of several classes of compounds.