Table 1.
Evaluation of isoform-level circular RNA prediction based on the 5-fold cross-validation
Method | Accuracy | Precision | Sensitivity | Specificity | F1-score | MCC | AUC |
---|---|---|---|---|---|---|---|
SVM | 0.7279 ± 0.0686 | 0.7479 ± 0.0970 | 0.8932 ± 0.1075 | 0.4526 ± 0.3413 | 0.8042 ± 0.0260 | 0.4031 ± 0.1784 | 0.6729 ± 0.1203 |
RF | 0.7607 ± 0.0084 | 0.7764 ± 0.0123 | 0.8610 ± 0.0077 | 0.5982 ± 0.0075 | 0.8165 ± 0.0095 | 0.4804 ± 0.0115 | 0.7296 ± 0.0053 |
Att-CNN | 0.7519 ± 0.0069 | 0.7739 ± 0.0257 | 0.8529 ± 0.0391 | 0.5872 ± 0.0527 | 0.8105 ± 0.0080 | 0.4612 ± 0.0165 | 0.7200 ± 0.0097 |
Att-RNN | 0.7638 ± 0.0075 | 0.7773 ± 0.0161 | 0.8582 ± 0.0345 | 0.6171 ± 0.0505 | 0.8152 ± 0.0094 | 0.4960 ± 0.0134 | 0.7377 ± 0.0105 |
nRC | 0.7557 ± 0.0115 | 0.7844 ± 0.0389 | 0.8410 ± 0.0597 | 0.6193 ± 0.0998 | 0.8094 ± 0.0118 | 0.4781 ± 0.0279 | 0.8280 ± 0.0091 |
PredcircRNA | 0.6550 ± 0.0076 | 0.6977 ± 0.0137 | 0.5949 ± 0.0070 | 0.7202 ± 0.0120 | 0.6422 ± 0.0091 | 0.3169 ± 0.0164 | 0.5882 ± 0.0102 |
circDeep | 0.8748 ± 0.0102 | 0.9393 ± 0.0134 | 0.8161 ± 0.0217 | 0.9407 ± 0.0138 | 0.8732 ± 0.0111 | 0.7584 ± 0.0186 | 0.7395 ± 0.0132 |
DeepCirCode | 0.8997 ± 0.0039 | 0.9353 ± 0.0228 | 0.9021 ± 0.0248 | 0.8967 ± 0.0383 | 0.9179 ± 0.0040 | 0.7914 ± 0.0073 | 0.8994 ± 0.0077 |
JEDI | 0.9878 ± 0.0007 | 0.9906 ± 0.0030 | 0.9906 ± 0.0032 | 0.9836 ± 0.0038 | 0.9904 ± 0.0009 | 0.9742 ± 0.0014 | 0.9872 ± 0.0009 |
Note: We report the mean and standard deviation for each metric.