Figure 5: Classification comparison of TAmC-Seq and MeDIP-Seq.
From: Tet-mediated covalent labelling of 5-methylcytosine for its genome-wide detection and sequencing
(a) The distribution of TAmC-seq and MeDIP-seq enrichment for a set of curated promoters that are consistently highly methylated (>75% mCG) within and around the CpG island/TSS, or consistently lowly methylated (≤25% mCG). (b) Random forest classification error in distinguishing methylated from unmethylated curated promoters in a. (c) Random forest classification error in distinguishing methylated from unmethylated promoters, for different CpG densities. An equal number of methylated and unmethylated promoters were randomly sampled, and subsequently classified for each CpG density class. Each line represents a different random sampling of promoters. The high variance of HCPs is due to the small number of methylated high-CpG promoters in HMEC. (d) A density plot of bisulphite change (%mCG(HCC1954)—%mCG(HMEC)) compared with TAmC change (RPKM(HCC1954)—RPKM(HMEC)) for promoters of various CpG content. The red line is the lowest weighted curve; RPKM, reads per million base pairs mapped per kb of model. (e) Random forest classification error in distinguishing hypermethylated from hypomethylated promoters for different CpG densities. An equal number of hypermethylated and hypomethylated promoters were classified for each CpG density class. Each line represents a different random sampling of promoters.
