Abstract
Long noncoding RNAs (lncRNAs) are increasingly recognized as druggable targets due to their conserved secondary/tertiary structures and regulatory roles in disease. A prototypical example is the MALAT1 triple helix, whose stability supports transcript persistence and is implicated in oncogenesis. Here, we evaluate the ability of a structure-based drug discovery (SBDD) pipeline, integrating molecular dynamics (MD), pocket analysis, ensemble docking, and diverse scoring functions, to capture the binding behavior of 21 congeneric diminazene-based ligands targeting MALAT1. Conformational ensembles were generated using both conventional MD and Hamiltonian Replica Exchange MD, revealing two potential binding sites. Ensemble docking with AutoDock GPU and rDock across representative RNA conformations, followed by rescoring with force-field and machine-learning-based scoring functions, led to the identification of a binding mode with the best agreement across the series. Principal component analysis of interaction fingerprints within clustered poses was used to explain the experimentally observed affinity trends. Our findings highlight the promise and limitations of current SBDD pipelines for flexible RNA targets and offer a benchmark for future improvement in RNA-focused drug discovery.