QC and Quantitation

Quality control

Trimming, mapping, FastQC and FastqQScreen reports were collated using MultiQC. WGBS and PCHiC sample quality was assessed with Bismark and HiCUP reports, respectively.

Post-mapping QC was performed with Seqmonk and Vistories assessing data distribution, data grouping, duplication, coverage, enrichment and any other specifics can be viewed here for each of the marks:

• TET ATAC
• TET CTCF
• TET H3K4me1
• TET H3K4me3
• TET H3K27ac
• TET H3K27me3

• TET RNA

• DNMT ATAC

• DNMT CTCF
• DNMT H3K4me1
• DNMT H3K4me3
• DNMT H3K27ac
• DNMT H3K27me3

• DNMT RNA

• DNMT and TET WGBS

• DNMT and TET HMCP

Excluded regions

ChIP input samples were used to exclude coverage outliers and some regions which were represented differently in DNMT and TET cell lines. Details can be found in this Seqmonk Vistory and the genomic locations (GRCm38) can be downloaded here.

Quantitation

PCHiC data as given in linear CHiCAGO scores (-log weighted p-values), if an interaction was found in both directions, the maximum score was used. Interactions were called significant and included in the analysis if their CHiCAGO score was ≥ 5 in any of the conditions as suggested by the method authors.

To integrate PCHiC with ChIP, ATAC and HMCP data, reads were quantitated across HindIII fragments and normalised for length and total number of reads in the sample (RPKM).

WGBS data was quantitated across 100 CpG using the Seqmonk methylation quantitation pipeline (min count per position = 1, min number of observations = 20) and is expressed in percent. To obtain methylation levels on HindIII fragment resolution, fragments were overlapped with 100 CpG windows. If a HindIII fragment overlapped with more than one 100 CpG window, the average was taken.

RNA data was quantitated using the Seqmonk RNA-Seq quantitation pipeline and is given as log2 RPKM values.

All values are replicate averages.

Data integration

PCHiC interacting pairs form the basis for UMAP and network representations. Quantitated data from different modalities were collated in a data table in which rows represent an interaction pair and columns are attributes of the interaction (PCHiC) or the respective bait and promoter interacting region (PIR) (ATAC, ChIP, HMCP, WGBS). Values are linear. For some visualisations (were level of the mark was represented as colour) capped values were used to obtain a more reasonable scale. Caps were introduced at 3x MAD (mean absolute deviation). Both bait and PIR were annotated with promoters (overlapping gene start coordinates); if more than one promoter was present in the HindIII fragment, the most upstream was used. Data tables for both raw and capped quantitations by interaction pair are provided as processed data on GEO.