Ensemble-aware computational approaches for RNA-targeted drug discovery

The expanding recognition of the central roles of RNA in cellular regulation is opening new therapeutic opportunities for diseases driven by its dysregulation. Two principal strategies are pursued in RNA based therapeutics, oligonucleotides and small molecules. While oligonucleotides have achieved clinical milestones, this thesis concentrates on small molecules, whose chemical tunability and compatibility with medicinal chemistry provide a complementary route. Despite growing interest, few marketed drugs directly target RNA, and most were identified through phenotypic screening. This thesis addresses three interconnected projects, progressing from comparative biophysical analysis to applied RNA targeting and experimental ensemble refinement. First, we compared ligand recognition in a protein enzyme and an RNA receptor that bind the same ligand, riboflavin kinase and the flavin mononucleotide riboswitch. Molecular dynamics simulations characterized conformational space, electrostatic organization, and ligand and solvent dynamics within binding pockets. The analysis revealed comparable capacities to establish polar noncovalent interactions, alongside distinct patterns of flexibility and solvent exchange. Second, we developed an early stage discovery framework targeting the HIV-1 trans activation response element. Mixed solvent simulations using the SHAdow Mixed solvent metAdynaMics approach generated hot spot and pocket maps that guided ensemble aware virtual screening. Conformational clustering was refined with local structural descriptors to preserve base pairing and stacking features. A hybrid pipeline integrating rigid receptor docking, a machine learning scoring function, and complementary chemical space exploration produced a shortlist of approximately two hundred candidate compounds. Third, we refined an HIV-1 TAR conformational ensemble by combining enhanced sampling with Maximum Entropy reweighting to reproduce residual dipolar coupling data. Well Tempered Ensemble simulations expanded conformational coverage, while reweighting enforced agreement with experimental averages. The refined ensemble enabled the characterization of dominant substates and their roles in ligand recognition. Together, these studies establish an ensemble centered framework for RNA targeted small molecule discovery.