Polypharmacology is usually represented as a relationship between one chemical structure and several targets, yet this representation suppresses a central physical complication: each target is itself an ensemble of interconverting conformations, and ligand recognition may occur through different subsets of those ensembles. Consequently, apparent multi-target activity can arise from different binding-competent states, different population shifts, and different kinetic pathways rather than from a transferable binding mechanism. This Perspective argues that molecular dynamics for polypharmacology should therefore compare competing conformational basins across targets rather than extend single-structure docking across a larger target panel. Evidence from G-protein-coupled receptors, nuclear receptors, kinases, enzyme isoforms, enhanced-sampling studies, and state-model methodology shows why static structural similarity, affinity, or a single predicted pose cannot by themselves establish cross-target conformational equivalence. We propose a cross-target ensemble logic in which target-local states are resolved first and compared only after their structural, thermodynamic, kinetic, and experimental support has been defined. This framing separates purposeful multi-target pharmacology from unresolved promiscuity while making uncertainty in pose selection, state definition, sampling, and transferability explicit.
IJPRAS
