A grammar-based approach to RNA pseudoknotted structure prediction for aligned sequences

A grammatical approach is proposed to predict RNA secondary structures including pseudoknots. The method is based on comparative sequence analysis, i.e., a prediction algorithm accepts a multiple alignment of RNA sequences. We use a stochastic multiple context-free grammar (SMCFG), which can precisely express a wide class of pseudoknots. The probability parameters for the SMCFG can be computed directly from aligned sequences. The experimental results show that the prediction performance of the proposed method is fairly high.