Skip to main content
Fig. 2 | Genome Biology

Fig. 2

From: ncHMR detector: a computational framework to systematically reveal non-classical functions of histone modification regulators

Fig. 2

Performance evaluation of ncHMR detector. a–c Graph showing robustness of cofactor identification on evaluation data with three types of noise that simulate three types of experimental variation. a A Gaussian noise on the H3K27me3 signal at EZH2 ChIP-seq peaks, the mean and standard deviation of the noise distribution were set to be equal to a given fraction (from 0 to 100%) of the average histone modification signal across the genome. b Random loss of a fraction (from 0 to 80%) of non-classical sites of EZH2. c Random alteration on a fraction (from 0 to 20%) of cobinding events with EZH2 for other factors. The F-beta score (β = 0.75) was used to evaluate the robustness by treating the factors in the strong negative group as true positive cofactors and other factors as true negative. The red, blue, yellow, and green lines represent cofactors identified by elastic net, Lasso, forward selection, and knockoff, respectively

Back to article page