A Classification System for Predicting RNA Hairpin Loops

Ribonucleic acid (RNA) molecules fold back over themselves to form secondary structures which determine the RNApsilas functionality in living cells. RNA secondary structure can be determined in laboratory by X-ray diffraction and nuclear magnetic resonance (NMR) techniques. However, these techniques are slow and expensive. Therefore, computational approaches are used to predict the secondary structure of RNA molecules. A new approach, RNA-SSP, for predicting RNA secondary structure elements is proposed. It combines computational approaches and machine learning classifiers to predict individual structure elements using a new search heuristic. The approach is implemented and tested for hairpin loops and a methodology for extending the approach to predict the remaining secondary structure elements is proposed. The experiments showed a significant improvement in prediction accuracy to 95% for stem regions and 80% for loops. The overall weighted-average accuracy for predicting hairpin loop sub-structure is 89% with a sensitivity of 85.29%.