Prediction of Pre-miRNA with Multiple Stem-Loops Using Feedforward Neural Network

摘要

MiRNA is a kind of single non-coding RNA that plays a pivotal regulated role in gene expression and has a very important influence in disease occurrence, growth and development, cell proliferation and so on. Therefore prediction miRNA has become the most important task in understanding miRNA regulation mechanism. Existing computational prediction methods are usually good at recognition pre-miRNA with multiple stem-loops. In this study, in order to further improve predictive precision of pre-miRNA, we quoted a set of new biologically multiple stem and loop secondary structure features based on the previous research work, then handled the imbalance problem of dataset, combined feedforward neural network. Finally, the new classifier system was constructed successfully with the proposed approach to separate the real pre-miRNA from datasest. By using the dataset of human pre-miRNAs and employing systematic 5-fold cross-validation methods for evaluating the classifier performance. We discover that the new classifier improved predictive precision effectively.