Application of advanced computational methods in drug discovery for targeting RNA

This thesis presents a novel computational approach to RNA-targeted drug discovery, addressing the challenges posed by RNA’s inherent flexibility and the limitations of traditional protein-docking protocols. The first part of the research focuses on two key aspects: druggability prediction and allosteric analysis. We introduce a one-class learning approach using the Import Vector Domain Description (IVDD) algorithm with customized DrugPred descriptors on pockets identified by NanoShaper. This method, validated on a dataset of 100 proteins from the Potential Drug Target Database (PDTD), offers a more nuanced and efficient approach to identifying druggable pockets compared to traditional binary classifications. While the investigation of allostery compares three computational methods – DyNet, DF, and Pocketron – across three pharmaceutical targets: the adenosine A2A receptor, androgen receptor, and EGFR kinase domain. Pocketron consistently demonstrates great performances in identifying known allosteric pockets with high correlation to the orthosteric site. Applying our refined protocols on proteins to the long non-coding RNA MALAT1, we used NanoShaper and Pocketron to identify potential target pockets that could disrupt the triple helix structure through long-range communication. Once the sites were defined we employ molecular dynamics simulations (unbiased and enhanced) to generate a comprehensive conformational ensemble. Having defined the ensembles we generated poses using two pose generation software (AutoDock GPU and rDock), we then evaluated various scoring functions (AutoDock, rDock, Vina, AnnapuRNA, and SPRank) for their ability to predict experimental binding affinities of diminazene-based ligands to MALAT1. Finally, we extend a non-equilibrium binding free energy estimation method to RNA molecules, focusing on the Riboswitch-preQ1 system. Using steered molecular dynamics and the Crooks Fluctuation Theorem, we calculate binding free energies for complexes with both cognate and synthetic ligands. This research contributes to the advancement of RNA-targeted drug discovery by providing novel computational tools and insights into the complex dynamics of RNA-ligand interactions.