Thesis (Ph.D.) - Indiana University, Informatics and Computing, 2016Highly chemically similar drugs usually possess similar biological activities but small changes in chemistry result in large differences in biological effects. Chemically similar drug pairs showing extreme deviations in activity represent distinctive drug interactions. The presence of these interactions adversely affects prediction of structure and activity associations. Their identification has crucial implications on drug development and innovations. Given the multitude of drugs in an ensemble, pairs possess multilevel distinctiveness in terms of their attributes of structural and activity similarity or variation. The cliff characterization for describing drops in similar activity has received considerable attention, however, it remains quantitatively less refined. In this dissertation, I investigate distinctiveness of drug interactions using a large drug-target network and provide a quantitative rationale for characterization of the pharmacological topography. I consider rises in pairwise similarity and variation in activity of drugs on proteins with chemical similarity (c) to assess levels of distinctiveness. These activity measures are affected by the presence of few drugs (targets) having multiple targets (drugs). I quantify interactions between drugs by considering similarity and variation jointly with c. The probability of distinctiveness is predicted by employing joint probability of structure and activity measures. Intermittent spikes in variation along the axis of c represent canyons in the activity landscape. This new representation accounts for distinctiveness through relative rises in activity measures and offers an enhanced perspective. It provides a mathematical basis for predicting the probability of occurrence of distinctiveness. It identifies the drug pairs at varying levels of distinctiveness and non-distinctiveness. Prediction is validated even if data approximately satisfy the conditions of the formulation. The difference in distinctive interactions emphasizes the importance of studying both measures, and reveals that the choice of measurement can affect the interpretation. Further, I find that minor changes in methods or perturbations of measures can crucially alter the classification of interactions as distinctive. Identification and interpretation of distinctiveness, therefore, gain relevance through methodological specifications. The present analysis of structure and activity provides an in depth modeling and assessment of distinctiveness and the probability of its occurrence. It could potentially influence decision-making in research and development.