Abstract
Many RNA molecules are dynamic, but characterizing their motions by experiments is difficult, often requiring application of complex NMR experiments. Computational methods such as molecular dynamics simulations, on the other hand, still suffer from difficulties in sampling and remaining force field errors. Here, we provide an atomic-level description of structure and dynamics of the 14-mer UUCG RNA stem-loop by combining molecular dynamics simulations with exact nuclear Overhauser enhancement data. The integration of experiments and simulation via a Bayesian/Maximum entropy approach enables us to discover and characterize a new state of this molecule, which we show samples two distinct states. The most stable conformation corresponds to the native, consensus three-dimensional structure. The second, minor state has a population of 11%, and is characterized by the absence of the peculiar non-Watson-Crick base pair between U and G in the loop region. By using machine learning techniques, we identify key contacts in the NOESY spectrum that are compatible with the presence of the low-populated state. Together, our results demonstrate the validity of our integrative approach to determine the structure and thermodynamics of conformational changes in RNA molecules.
