Abstract:

This paper improves the MiRfilter system by parameter, range of parameter and score in mature miRNAs in order to predict miRNAs in plants. We predict miRNAs in Populus trichocarpa genome and use its positive and negative samples to examine the accuracy of MiRfilter’s prediction. MiRfilter predicts 3 860 Populus trichocarpa miRNA candidates in all. It correctly identifies 91 precursors and 80 matures in 110 positive samples. Accuracies of the precursor and mature prediction reach 82.73% and 72.73%, respectively. We find 12 false positive miRNAs from 1 968 negative samples in LG_Ⅳ chromosome, and the false positive rate reaches 0.61%.

Key words: plant miRNA; MiRfilter; Populus trichocarpa genome; oneclass methods;  K nearest neighbor (KNN)