Drug discovery is both a long and expensive process, characterized by low success rates and high costs of development. By identifying new therapeutic applications for existing drugs, drug repurposing represents another faster and less expensive alternative using safety profiles and pharmacokinetic data that are already established. Nevertheless, the drug repurposing field faces obstacles such as data scarcely integrated with one another, a lack of insight into molecular mechanisms, and difficulties in the integration of different types of such data. Chemoinformatics addresses the gaps of repurposing drug information by employing methods such as ligand- and structure-based virtual screening, molecular docking, and pharmacophore modeling. A number of tools are available for identifying drug–target interactions, making a shift toward a polypharmacological perspective. The use of three-dimensional molecular descriptors enables more accurate screening, mainly accounting for the molecular conformation and complex interactions. On the other hand, machine learning and deep learning, by using large amounts of data, help to predict drug–target interaction and new therapeutic uses on an unprecedented scale. Recent advances, such as AlphaFold for protein folding and more recently interaction prediction, increase the accuracy of drug repurposing while accelerating the candidate hit discovery timelines. In this review, we highlight several chemoinformatics and machine learning approaches used for different drug development-related tasks and discuss how these approaches can guide drug repurposing to tackle complex diseases and rapidly address emerging health crises.
Sirci et al. (Wed,) studied this question.