Computational Suite for RNA-Targeted Drug Design

DESCRIPTION (provided by applicant): We aim to extend and validate computational methods and protocols for treating accurately the interactions among RNA, ions and ligands. Ultimately, we seek tools that are efficient, accurate and useful in a drug discovery framework for identifying small-molecule drugs for RNA therapeutic targets. As research continues to reveal ever-expanding roles for RNA in biology, functional RNA molecules are recognized increasingly as attractive drug targets. Moreover, functional RNAs, like protein targets, display sequence-specific motifs and adopt functional three-dimensional folds. Comparatively few computational studies have focused on RNA systems. This imbalance is due in large part to the historical lack of high-resolution structural and associated biochemical data for systems beyond very small model oligonucleotides. In fact, the high-resolution crystal structures of the SOS ribosomal subunit of Haloarcula marismortui (H50S) and of several known antibiotics complexed with this H50S that have been solved recently by our founders and our group at Rib-X represent a huge advance in the information available for such studies. Importantly, the antibiotics principally and nearly exclusively interact with the nucleic acids and ions in the SOS ribosomal subunit; that is, protein ligand interactions are largely absent in these crystal structures. We have identified three areas of emphasis for treating efficiently and accurately RNA systems: (1) the development of a computational approach to identify ion binding sites and preferences in RNA model systems, (2) the exploration of nucleic acid flexibility in simulating RNA systems and (3) the identification of an energy scoring function for explaining and predicting the binding of small molecules to RNA systems. Our overall theoretical approach emphasizes fully-atomistic calculations and simulations using BOSS and NICPRO, which allow molecular mechanics calculations and Monte Carlo simulations, and ANALOG(tm) , which is a growing program for scoring RNA-ligand interactions. For the treatment of ions, we will investigate a variation of multiple-copy simulation techniques, employing Monte Carlo simulation. We will look for optimal location, orientation and type of ions in a diverse training set of RNA systems where ions are known to influence structure and function. Solvent will be treated implicitly using a Generalized-Born approach. Optimally, we will attempt to identify general patterns of ion binding in RNA systems, with an emphasis on binding sites or functionally-relevant sites. The simulation results will be compared with structural and corroborating biochemical data. Success will be judged on the ability to identify correctly the most favorable ion binding spots and the type of ion observed to populate those sites. To assess nucleic acid flexibility and anticipate it generally, an extension of a newly-implemented concerted rotation algorithm will be performed, both in the molecular mechanics calculations and in the Monte Carlo simulations. The modification will allow us to sample the backbone, the conformationally-promiscuous sugar 2'-OH groups and sugar repuckering. Fitness of this approach will be judged on the ability to reproduce significant crystallographically-observed conformational changes of nucleic acids in H50S upon the binding of a family of small-molecule ligands. A rapid and accurate approach is needed in the drug discovery world for rank-ordering small-molecule compounds for activity and affinity with RNA drug targets. Our scoring tools will be enhanced with a protocol that incorporates a method for ion representation and nucleic acid flexibility identified in this research program. The modified ANALOG(tm) will then be used in a retrospective study to identify key molecular properties (descriptors) that contribute to intrinsic affinity in each of three ri bosom ally-targeted programs, two SOS-targeting and one SOS-targeting. Standard regression models will be attempted, and statistical measures will be monitored, focused on the goodness of fit (r2 and q2), the significance of individual molecular descriptors and the variance in the models.