We organized this Research Topic to bring together work that uses molecular modeling, broadly defined, to advance drug repurposing. The collection includes five articles covering pharmacophore design, transporter biology, network pharmacology, transcriptomics-guided docking, and RNA-protein targeting. Together, they reflect how diverse the computational toolkit for repurposing has become. Elsaka et al. review pharmacophore modeling, from its origins as a theoretical concept to its current use as a routine virtual screening tool. Their discussion of dynamic pharmacophore models (dynophores) derived from MD trajectories is particularly relevant, as these models account for binding-site flexibility that static approaches miss. The review also covers recent applications of AI to feature extraction and hit-rate improvement, illustrated with case studies on efflux pumps, topoisomerase II-alpha, and LEDGF/p75integrase inhibitors.In a different therapeutic context, Kaijage and Kraszewski focus on sodium-glucose cotransporters (SGLTs) and the challenge of achieving selective SGLT1 inhibition. They bring together cryo-EM structures, AlphaFold2 predictions, and free energy calculations from MD simulations to propose a roadmap for designing selective inhibitors. The review highlights how structural data from different sources can be combined computationally to guide rational drug design for metabolic and cardiovascular targets.Hu et al. take a data-driven approach to intestinal ischemia-reperfusion injury, using WGCNA and machine learning to identify five mitochondrial-related hub genes (Pdk4, Yrdc, Bcl2l11, Bcl2a1d, and Pmaip1). Molecular docking serves here as a validation step rather than the primary discovery method, linking the transcriptomic findings to potential therapeutic compounds, specifically securinine and ABT-737. This work illustrates how docking can complement omics-based pipelines even when it is not the driving methodology.Prakash addresses brain cancers, a setting where tumor heterogeneity makes drug selection particularly difficult. The study builds molecular profiles from signaling pathway components (EGFR, BRAF, PDGFRA, TP53, CDKs) and screens 2,809 FDA-approved drugs through two purpose-built tools: "in-mac" for profiling and "ReBrain" as a network-based database. The framework reports 70-95% accuracy and identifies mefloquine, clofibric acid, and armillarisin A as priority candidates, showing that network pharmacology can handle the complexity of heterogeneous tumors. Finally, Smith et al. push the boundaries of what counts as a druggable target. Their perspective proposes ribonucleoprotein (RNP) interfaces in viral 5' untranslated regions as sites for small-molecule intervention, using Enterovirus A-71 as a proof of concept. The workflow combines structural prediction, ensemble-based MD simulations, virtual screening, and biophysical validation. Targeting RNA-protein interaction surfaces for repurposing remains largely unexplored, and this contribution lays out a concrete strategy to address it.Taken together, these articles point to a few recurring observations that we believe are worth highlighting. Modern repurposing campaigns rarely rely on a single computational method. Instead, the most convincing results come from combining docking, MD, network analysis, or ML in pipelines where each technique compensates for the limitations of the others. At the same time, the structural and dynamic characterization of drug-target interactions remains the foundation that gives computational predictions their mechanistic grounding, whether the target is a protein, a transporter, or an RNA-protein complex. Equally important, none of these computational strategies can stand alone without experimental validation to confirm that in silico predictions translate to measurable biological effects.
Filipe et al. (Wed,) studied this question.