Predicting RNA secondary structure based on the class information and Hopfield network

One of the models for RNA secondary structure prediction is to view it as maximum independent set problem, which can be approximately solved by Hopfield network. However, when predicting native molecules, the model is not always accurate and the heuristic method of Hopfield network is not always stable. It is because that the class information is lost and the accuracy is not determined by the number of base pairs only. Secondary structures of non-coding RNAs are believed conservative on the same class. However, software and web servers nowadays for RNA secondary structure prediction do not consider the class information. In this paper, we involve class information in the initialization of Hopfield network. When the initialization is improved, the promising experimental result shows the efficacy and superiority of our proposed methods.