A fuzzy dynamic programming approach to predict RNA secondary structure

Due to the recent discovery of many RNAs with great diversity of functions, there is a resurgence of research in using RNA primary sequences to predict their secondary structures, due to the discovery of many new RNAs with a great diversity of functions. Among the proposed computational approaches, the well-known traditional approaches such as the Nussinov approach and the Zuker approach are essentially based on deterministic dynamic programming, whereas the stochastic context-free grammar (SCFG), the Bayesian estimation, and the partition function approaches are based on stochastic dynamic programming. In addition, heuristic approaches like artificial neural network and genetic algorithm have also been presented to address this challenging problem. But the prediction accuracy of these approaches is still far from perfect. Here based on the fuzzy sets theory, we propose a fuzzy dynamic programming approach to predict RNA secondary structure, which takes advantage of the fuzzy sets theory to reduce parameter sensitivity and import qualitative prior knowledge through fuzzy goal distribution. Based on the experiments performed on a dataset of tRNA sequences, it is shown that the prediction accuracy of our proposed approach is significantly improved compared with the BJK grammar model of the SCFG approach.