Probing Local Atomic Environments to Model RNA Energetics and Structure

Abstract

Ribonucleic acids (RNA) are critical components of living systems. Understanding RNA structure and its interaction with other molecules is an essential step in understanding RNA-driven processes within the cell. Experimental techniques like X-ray crystallography, nuclear magnetic resonance (NMR) spectroscopy, and chemical probing methods have provided insights into RNA structures on the atomic scale. To effectively exploit experimental data and characterize features of an RNA structure, quantitative descriptors of local atomic environments are required. Here, I investigated different ways to describe RNA local atomic environments. First, I investigated the solvent-accessible surface area (SASA) as a probe of RNA local atomic environment. SASA contains information on the level of exposure of an RNA atom to solvents and, in some cases, is highly correlated to reactivity profiles derived from chemical probing experiments. Using Bayesian/maximum entropy (BME), I was able to reweight RNA structure models based on the agreement between SASA and chemical reactivities. Next, I developed a numerical descriptor (the atomic fingerprint), that is capable of discriminating different atomic environments. Using atomic fingerprints as features enable the prediction of RNA structure and structure-related properties. Two detailed examples are discussed. Firstly, a classification model was developed to predict Mg$^{2+}$ ion binding sites. Results indicate that the model could predict Mg$^{2+}$ binding sites with reasonable accuracy, and it appears to outperform existing methods. Secondly, a set of models were developed to identify cavities in RNA that are likely to accommodate small-molecule ligands. The models were also used to identify bound-like conformations from an ensemble of RNA structures. The frameworks presented here provide paths to connect the local atomic environment to RNA structure, and I envision they will provide opportunities to develop novel RNA modeling tools.