Coincidence algorithm for RNA secondary structure prediction with Pseudoknots

Pseudoknots are special units in RNA structure. They are found in almost all categories of RNAs. Since the prediction of the pseudoknotted structure has been proven to be NP-complete, heuristic methods have been proposed to alleviate the issue of computational complexity. The Coincidence algorithm (COIN) was applied to predict the RNA secondary structure with pseudoknots. The proposed method relaxed some conditions to identify structures that contain pseudoknots, and also presented a new fitness function that combines approximate minimum free energy (MFE) and maximum expected accuracy (MEA). The new fitness function makes the algorithm faster than using standard MFE but still reflects the quality of the predictive structure. The proposed method was tested on the Pseudobase ++ database and compared with the PRSA based on Simulated Annealing (SA), IPknot, TT2NE, and RNAstructure. On average, the proposed method outperformed the other methods. It yields F1 scores 3 % to 24 % higher than the compared methods.