Table 6. Comparison of methods’ sensitivity for tests performed with the CLIP-Seq data.
Method | Features/Setup | Data set #1 | Data set #2 |
RF | complete set | 0.704 | 0.725 |
top 12 | 0.774 | 0.756 | |
SVM | complete set | 0.464 | 0.549 |
top 12 | 0.901 | 0.891 | |
NB | complete set | 0.591 | 0.657 |
top 12 | 0.633 | 0.689 | |
KNN | complete set | 0.661 | 0.756 |
top 12 | 0.675 | 0.633 | |
J48 | complete set | 0.521 | 0.586 |
top 12 | 0.732 | 0.743 | |
GLM | complete set | 0.000 | 0.027 |
top 12 | 0.000 | 0.018 | |
TargetSpy | seed | 0.421 | 0.339 |
no-seed | 0.459 | 0.529 |
Data set #1 refers to interactions predicted by all softwares except miRanda (38 pairs).
Data set #2 comprises interactions predicted by all softwares, including miRanda (170 pairs).
Both TargetSpy tests were performed using the sensitivity as the threshold.