Fig. 2
From: Predicting enhancers in mammalian genomes using supervised hidden Markov models
Validation. a Precision recall curves of eHMM in within and across sample validation schemes on the FANTOM5 data in mouse ESC, liver and lung. Circles indicate prediction performance of the viterbi algorithm, while the lines represent precision and recall based on posterior probabilities obtained from the forward-backward algorithm. b-c Comparison of areas under precision recall curve using different enhancer prediction methods validated on regions from FANTOM5 (b) and Enhanceratlas (c). Legend acronyms: CV - within-sample 5-fold cross-validation. ESC - across-sample validation using a model trained on ESC data including quantile normalization. ESC raw - across-sample validation using a model trained on ESC data without normalization. n - number of states