Table 3.
Comparison of ACP predictors trained and tested using the alternative data set. Results were obtained from AntiCP2.0 [11] and ACPred [10], except AI4ACP.
| Predictors | Classifier | Accuracy | Sensitivity | Specificity | MCC * |
|---|---|---|---|---|---|
| AntiCP | SVM | 0.900 + | 0.897 | 0.902 | 0.800 |
| iACP | SVM | 0.776 | 0.784 | 0.768 | 0.550 |
| ACPred | SVM | 0.853 | 0.871 | 0.835 | 0.710 |
| PEPred-Suite | ensemble approach | 0.575 | 0.402 | 0.747 | 0.160 |
| ACPred-FL | ensemble approach | 0.438 | 0.602 | 0.256 | −0.150 |
| ACPred-Fuse | RF | 0.789 | 0.644 | 0.933 | 0.600 |
| AntiCP2.0 | ETree | 0.920 | 0.923 | 0.918 | 0.840 |
| iACP-FSCM | SVM | 0.889 | 0.876 | 0.902 | 0.779 |
| AI4ACP | CNN | 0.894 | 0.871 | 0.918 | 0.790 |
*: Matthews Correlation Coefficient. +: Top two ranked methods for each index are presented using text formats: first in boldface, second with underline.