Skip to main content
. 2014 Jul 25;42(15):10086–10098. doi: 10.1093/nar/gku681

Table 5. Summary of the independent benchmark RB67 results in terms of MCC.

Evaluation Method Accuracy Specificity [+] (Precision) Sensitivity (Recall) F-measure MCC AUC
aaRNA 0.882 0.437 0.494 0.464 0.399 0.842
BindN+ 0.862 0.372 0.491 0.423 0.351 0.814
Residue- RNAbindR 2.0 0.867 0.376 0.438 0.404 0.331 0.798
based Seq-CTRL 0.886 0.443 0.401 0.421 0.358 0.811
KYG 0.804 0.274 0.542 0.364 0.284 0.780
DRNA 0.842 0.298 0.392 0.339 0.254 N/A
OPRA 0.843 0.301 0.403 0.345 0.261 N/A
aaRNA 0.844 0.428 0.449 0.398 0.323 0.814
BindN+ 0.828 0.377 0.463 0.397 0.301 0.780 *
Protein- RNAbindR 2.0 0.750 0.296 0.616 0.372 0.272 0.764 **
based Seq-CTRL 0.797 0.355 0.488 0.372 0.286 0.787 ***
KYG 0.769 0.298 0.505 0.349 0.240 0.716 ****
DRNA 0.795 0.319 0.397 0.331 0.229 N/A
OPRA 0.797 0.242 0.259 0.203 0.116 N/A

Different predictors were compared in the same way as the RB44 benchmark. When tested on this up-to-date benchmark, the aaRNA got a superior performance than all others. Paired Wilcoxon tests on protein-averaged AUCs of aaRNA and other methods indicated significant differences (P* < 2e-4, P** < 2e-5, P*** < 7e-5 and P**** < 4e-9).