Skip to main content
. 2014 Mar 7;42(9):e76. doi: 10.1093/nar/gku182

Figure 2.

Figure 2.

Identification of confidence or validated targets. (A) Based on positive/negative targets and the prediction scores, ROC and PRC were generated for the six methods, where Seed Matrix, TargetScan Context Score and PCT are sequence-based and the other three are proposed competition models. (B) The number of validated targets selected by each method among their top 1000 rankings as a function of sample. Ten panels correspond to 10 cancer types. The curves are smoothed using loess from R function geom_smooth. (C) Comparison of the competition models with other expression-based methods including Pearson correlation, LASSO and GenMiR++.