We report a new ribonucleic acid （RNA） base discrete state model, which was first developed in our lab and designed to provide an efficient and accurate way of representing RNA structures toward RNA three-dimensional structure predictions. Since RNA free energy is largely determined by base pairs and base stackings instead of backbone trajectories, we directly model the RNA base configurations with respect to its previous one along the sequence. This is in sharp contrast with all previous works where the backbone trace was represented. To test how faithfully the discrete model can reproduce the chain trace in continuous space, we randomly select partial chains from the native structure of 23S ribosome RNA and re-grow them. The rms distance of the re-grown structures from the native ones is ～1.7 ？ for an optimized 16-state discrete model and gradually increases to ～3.3 ？ for long chains of length 50. The efficiency is also good, e.g. the program will finish within several tens of second for long loops of length 50. Our model may facilitate the RNA three-dimensional structure predictions in the near future when combined with appropriate free energy evaluation methods.
In this work, the traditional method of potential of mean force (PMF) is improved for describing the protein-protein interactions. This method is developed at atomic level and is distance-dependent. Compared with the traditional method, our model can reasonably consider the effects of the environ- mental factors. With this modification, we can obtain more reasonable and accurate pair potentials, which are the pre-requisite for precisely describing the protein-protein interactions and can help us to recognize the interaction rules of residues in protein systems. Our method can also be applied to other fields of protein science, e.g., protein fold recognition, structure prediction and prediction of thermo- stability.
