含伪结的RNA分子二级结构预测

RNA secondary structure prediction with pseudoknots

预测含伪结的RNA分子二级结构是生物信息学的一个研究难点。利用多分类支持向量机结合贝叶斯神经网络针对含伪结的RNA分子二级结构进行预测。利用多分类支持向量机进行预测,输出端得到相应碱基的平面伪结结构的E-NSSEL(Ex-tend New Secondary Structure Element Label)类别标签。使用碱基已预测的结果通过贝叶斯神经网络进行修正,并恢复RNA分子二级结构。使用该方法能有效地改善含伪结的RNA分子二级结构的预测效果。

RNA secondary structure prediction with pseudoknots is one of the most difficult research areas in bioinformatics.This paper introduces a new representation of the RNA secondary structure with plane pseudoknots by multi-class Support Vector Machine（multi-class SVM） and Bayesian Neural Networks（BNN）.A multi-class SVM model is presented to predict RNA secondary structure based on E-NSSEL labels that can express plane pseudoknots effectively.BNN is used to correct the results by considering the neighbor residues predicted labels.The RNA secondary structure is resumed according to the predicted results.Experiment proves that this method can improve the RNA secondary structure prediction results with plane pseudoknots.