An Intuitionistic Fuzzy Set Analysis of Drug-Target Interactions

The use of drugs in modern medicine is widespread and no doubt has played a significant role in improving outcomes in the treatment of many medical conditions. Hence, the discovery of drugs for new therapeutic indications as well drugs of enhanced efficacy for existing indications are of paramount importance. Information on drug-target interactions plays an important role in this task its importance is reflected by the fact that searching the Internet based on 'drug-target interaction' yields 280,000,000 hits. The interaction of a given drug with respect to multiple drug targets also known as polypharmacology is becoming increasingly important in modern drug research as it bears on the role of the biological pathways underlying a drug's mechanism(s) of action, the existence of side-effects, and the repurposing of existing drugs for new therapeutic indications. The ability of multiple, structurally dissimilar drugs to interact with a given target, also known as polyspecificity, is the complement of polypharmacology, and while not as well-known has nevertheless played a role in drug discovery research under the rubric of multiple lead series. Both of these concepts are considered in this work. Although sets of n drugs and m targets give rise to n X m virtual drug-target pairs, experimental and computational interaction data typically exist for only a small subset of these pairs. Thus, polypharmacology and polyspecificity values will, in all likelihood, be underestimated. As is shown in this work, taking account of drug-target pairs of unknown interaction (i.e. 'null' pairs) yields an upper bound to these values. However, in order to include such information in their analysis it is desirable that a methodology be able to handle null pairs in a reasonably straightforward manner. This is not the case with classical sets where the interaction of drug-target pairs is typically represented by set-theoretic relations, where a '1' indicates the presence of an interactive drug-target pair- with respect to a given interaction threshold and a '0' indicates the presence of a non-interactive pair. A new set-theoretic formalism based on extreme-value intuitionistic fuzzy sets (IFS) is utilized in this work since it explicitly accounts for the uncertainty inherent in null drug-target pairs. As will be demonstrated in future work, it can be generalized to include drug-target pairs with real-valued interactions. A simple example dataset is provided illustrating the application of IFSs to the analysis of drug-target datasets.