Skip to main content
. 2018 Sep 7;14(12):1669–1677. doi: 10.7150/ijbs.27819

Table 1.

Prediction results of different classifiers via cross validation.

Species1 Classifiers2 Acc3 Sn3 Sp3 MCC3 AUC3 AUC013
Mammalia
full transcript		 RFENAC 86.13 47.12 90.02 0.314 0.806 0.0340
RFKmer 85.39 37.33 90.18 0.241 0.769 0.0255
RFKSNPF 85.39 34.45 90.46 0.222 0.769 0.0219
RFPseDNC 85.17 30.70 90.60 0.193 0.727 0.0204
UGRU 87.48 62.21 90.00 0.423 0.885 0.0403
BGRU 87.57 63.15 90.00 0.430 0.889 0.0413
BERMP 87.80 65.76 90.00 0.448 0.891 0.0456
Mammalia
mature mRNA		 RFENAC 85.74 38.80 90.42 0.256 0.761 0.0251
RFKmer 84.38 22.79 90.58 0.125 0.666 0.0143
RFKSNPF 83.75 20.08 90.09 0.094 0.623 0.0132
RFPseDNC 84.00 19.80 90.40 0.095 0.621 0.0124
UGRU 85.90 43.73 90.10 0.289 0.813 0.0263
BGRU 85.90 44.74 90.00 0.296 0.815 0.0272
BERMP 86.14 46.58 90.08 0.311 0.817 0.0294
Saccharomyces cerevisiae
mRNA		 RFENAC 67.64 44.91 90.36 0.396 0.792 0.0285
RFKmer 61.41 32.27 90.55 0.281 0.724 0.0207
RFKSNPF 60.45 30.27 90.64 0.262 0.719 0.0209
RFPseDNC 58.77 27.36 90.18 0.226 0.693 0.0153
UGRU 54.86 19.45 90.27 0.138 0.648 0.0101
BGRU 56.86 23.64 90.09 0.184 0.679 0.0142
BERMP 68.59 47.10 90.10 0.412 0.800 0.0280
Arabidopsis thaliana
mRNA		 RFENAC 81.02 71.71 90.33 0.632 0.898 0.0511
RFKmer 85.53 81.24 90.02 0.714 0.928 0.0612
RFKSNPF 84.33 78.67 90.00 0.691 0.919 0.0572
RFPseDNC 83.55 77.10 90.00 0.677 0.910 0.0514
UGRU 84.95 79.71 90.20 0.703 0.923 0.0581
BGRU 85.93 81.71 90.14 0.721 0.928 0.0583
BERMP 85.95 81.81 90.10 0.722 0.927 0.0582

Note: 1 The datasets and the number of folds for cross validation were depicted in Figure S1. 2RFENAC=RF classifier with the ENAC encoding, RFKSNPF= RF classifier with the encoding of K-spaced nucleotide pair frequencies, RFPseDNC=RF classifier with the encoding of Pseudo dinucleotide composition, UGRU= the unidirectional GRU-based RNN classifier with word embedding, BGRU= the bidirectional GRU-based RNN classifier with word embedding, BERMP= BGRU-based Ensemble RNA Methylation site Predictor that integrating BGRU and RFENAC. 3Acc=accuracy, Sn=sensitivity, Sp=specificity, MCC=Matthew's Correlation Coefficient, AUC=area under the receiver operating characteristic, AUC01 = AUC with a <10% false positive rate (i.e., specificity>90%).