RNA Secondary Structure Prediction With Plane Pseudoknots Based On Support Vector Machine

With the development of Human Genome Project implementation and the rapid growth of biological data information, it is a huge challenge for computer science to extract useful information from lots of data and reveal the biological meaning that the data contains. To make clear biological molecular functions in the process of life from its structure, structure prediction should be an essential question. Pseudoknot prediction is one of the most difficult researches in RNA secondary structure prediction areas. Current algorithms mainly predict plane pseudoknots. Plane pseudoknot playes a key role in the process of RNA functions realization. It has great pratical significances to predict RNA secondary structure with plane pseudoknot.Firstly, this paper studies for the problem of RNA bases predicted classification,it gives a predicted classification algorithm for RNA bases classified based on FDAGSVM. Exact classification of RNA bases is the premise of correct prediction on RNA secondary structure. The algorithm carries on fuzzy processing to decision-making results, and calculates the fuzzy membership instead of checking symbolic operation, then chooses the decision-making path in the directed acyclic graph based on fuzzy membership of the sample.Secondly, to solve the coding problem which is in the process of restoring RNA secondary structure based on E-NSSEL label sequences, this paper proposes a base pair coding method that the two bases which might get paired in the RNA sequences encoded as a whole. The method evinces a base local environment model with vectors firstly, and then connects the local environment model vectors of the two bases which might get paired to form the base pair vectors. Lastly, by using the E-NSSEL label sequences of the RNA sequences that obtained by front predicted, another two-class SVM is used to judge and restore the RNA secondary structure according to the predicted results.