Journal de protéomique et bioinformatique
Libre accès

Abstrait

HTar: Hidden Markov Model Based MicroRNA Binding Site Prediction

Salim A and Vinod Chandra SS

MicroRNAs are small, non-coding RNA molecules that regulate gene expression. MicroRNA may binds to mRNAs and control the intended function of mRNAs. There are a handful of computational algorithms for target prediction, but the degree of false positives and false negatives are high. In this paper, we propose a Hidden Markov Model for seed prediction and a Support Vector Machine (SVM) classifier for target prediction. Positive data set for training has been collected from experimentally validated targets, while negative data set has been identified systematically from predicted false positives. Each mRNA target candidate sequence is aligned with microRNA sequence and tested for a seed region using the trained HMM model. If the test succeeds, 22 features were extracted from the aligned duplex and fed into an SVM classifier. HMM based seed identification module works with an accuracy of 95.6% and SVM classifier provides 97.49% accuracy. We have compared binding sites of 9 microRNA in 148 m RNAs with the results of validated target sites and our results are more accurate than other approaches.