Optimal Use of Conservation and Accessibility Filters in MicroRNA Target Prediction
Figure 4
Combination of conservation and accessibility can improve performance over individual filters.
(A–B) Number of validated targets per miRNA (A) and precision (B) of the topmost predictions of PACCMIT for highly conserved miRNAs. (C–D) Comparison of PACCMIT (using the combined filter) with different methods under the same conditions as in panels (A–B). In panel (D), the precision of the top two predictions per miRNA is not defined for MirSVR. (E) Number of validated targets found before the first, second, and third false positive appears in the ranked predictions for highly conserved miRNAs. For this analysis we used the ‘large’ dataset of validated targets that involve all 74 highly conserved miRNAs (see Methods).