RNA structure prediction is one of the major topics in bioinformatics. Among the various RNA structures, pseudoknots are the most complex and unique structure. Various methods have been used for modeling RNA pseudoknotted secondary structure. In this paper a new model for prediction of RNA pseudoknot structure has been proposed. In this model, features of two existing techniques, i.e. neural network and grammar are combined. The advantage of grammar, identification based on rules is combined

more »

... h the strength of a neural network to learn. An Elman neural network is used to learn the context free grammar that represents a pseudoknot. This Learning grammar network further identifies if the RNA sequence contains pseudoknot or not. Learning grammar helps in reducing the drawbacks of both neural network and grammar thus increasing the overall power of identifying sequences with pseudoknots. Highlights  Pseudoknots are the most complex and unique structure. It is very difficult for an algorithm to identify all the classes of pseudoknots at once.  In the proposed Learning Grammar Model for Pseudoknot Identification features of existing technology neural network and grammar are combined.  In this model Elman neural network is used. Elman neural network tries to learn context free grammar that represents a pseudoknot. After learning the neural network can classify RNA sequence into sequences with or without pseudoknots.  For sequences with pseudoknots the model could detect 80.34% of the sequences. It could detect all of the sequences without pseudoknots correctly.  Combining neural network with grammar helps in reducing the drawbacks of both the technologies and increasing the overall power of identifying sequences with pseudoknots