A new application using approximate similarity (AS) measurements to the study of chemical properties of drugs is presented in this paper. A quantitative structure-activity relationship (QSAR) model for predicting spirosuccinimide fused tetrahydropyrrolo[1,2-a]pyrazine (SPPP) compounds activity as inhibitors of the aldose reductase (AR) enzyme was developed. This enzyme is involved in the transformation of glucose into sorbitol, which causes several diseases related to diabetes mellitus. AS matrices were built based on isomorphic and nonisomorphic data fusion and they were employed as representation space of SPPP data set for the development of QSAR model. For this purpose, isomorphism among all the pairs of molecules of the studied data set was extracted and Wiener and HyperWiener descriptors were used for describing the isomorphic and nonisomorphic subgraphs. Full cross validation (Q² = 0.92, Standard Error in Cross Validation = 0.10) was the strategy employed to build and validate accurate predictive similarity spaces.