Computational Features Evaluation for RNA Secondary Structure Prediction

Computational prediction of RNA secondary structure is classical open problem in computational molecular biology. Comparative sequence analysis is the gold standard method when given homologous sequences alignment. The essential of this method is a classification problem: to judge if any two columns of an alignment correspond to a base pair using provided information by alignment. However, all existing prediction methods select computational features by qualitative analysis, without a uniform criterion. Here, we collected various computational features used in existing prediction methods, and quantitatively compare the classification capability of those features by feature selection technique. As a result, an optimum subset of features was selected for predicting RNA secondary structure by classification. The test on 49 Rfam alignments shows the effectiveness of the selected features.