Fig. 6.

Analyses of the relations between PEPMAN prediction scores and different regulatory factors, including (A) histone modifications, (B) transcription factor binding sites, and (C) DNA methylation. In A and B, negative control was generated by randomly sampling the genomic regions from the same genes, each of which did not overlap with any binding area of the corresponding epigenetic factor and had the same length of an epigenetic binding peak. The sums of PEPMAN prediction scores of individual sites within peak regions were compared. In C, 5,000 randomly selected positions in the test data in which the guanine–cytosine content distributions were consistent with the DNA methylation sites were used as background. The PEPMAN prediction scores of the DNA methylation sites were normalized by the median of the prediction results of background. *: $1\times 10^{-5}<P<0.05$, two-sided Wilcoxon rank-sum test; **: $1\times 10^{-50}<P<1\times 10^{-5}$, two-sided Wilcoxon rank-sum test; ⁺: $P<1\times 10^{-50}$, two-sided Wilcoxon rank-sum test; ^-: $P>0.05$, two-sided Wilcoxon rank-sum test.