Ryder: Epigenome normalization using a two-tier model and internal reference regions

Key results

1. ShownIn DNase-seq of Gata3 knockout (KO) versus wild-type DN3 thymocytes, reads-per-million (RPM) normalization showed a global signal rise even at invariant CTCF-associated DNase hypersensitive sites (DHSs). Ryder aligned these reference regions and revealed the expected loss of accessibility at GATA3-bound DHSs.

   KO-increased GATA3-bound DHSs fell from 687 under RPM to 8 after Ryder; KO-decreased ones rose from 73 to 320. S3norm left reference DHSs significantly lower in KO and missed the GATA3-bound differences. Boxplot density comparisons used two-sided paired Wilcoxon signed-rank tests across matched regions; the site counts come from a fold-change ≥ 2 threshold on MA plots. · 3.1 Ryder normalizes epigenomic signal tracks…; 3.2 Ryder corrects technical signal inflation in…; Figure legends

2. ShownAcross three BRG1-loss datasets (auxin-inducible degron DNase-seq in mouse fibroblasts, dTAG13 titration CUT&RUN and ATAC-seq in mouse embryonic stem cells (mESCs), and inhibitor-treated human MV411 ATAC-seq), Ryder showed enhancer-preferential loss of BRG1 binding or accessibility, with promoters less affected.

   In the dTAG series, RPM-normalized CUT&RUN carried a background shift, whereas Ryder showed dose-dependent loss of BRG1 binding and of ATAC-seq accessibility at enhancers; promoter binding changed only under extensive degradation. In MV411 (AU15330, BRM014), RPM gave a spurious promoter accessibility gain. S3norm missed or weakened the enhancer-specific pattern in all three datasets. · 3.4 DNase-seq analysis of BRG1-AID cells with…; 3.5 Ryder recovers BRG1-dependent enhancer…; Figure legends

3. SuggestedRyder kept expected global shifts rather than flattening them. H3K27me3 fell after EZH2 inhibition without a spurious H3K4me3 gain, H3K9ac rose after HDAC inhibition, and BRM014 raised MNase-seq nucleosome occupancy at enhancer centres.

   In PC9 cells (DMSO vs GSK126), RPM normalization produced an apparent H3K4me3 increase that Ryder, using peaks overlapping invariant CTCF sites, avoided. In HeLa-S3 cells (DMSO vs Trichostatin A, with variable yeast spike-in ratios), Ryder showed the anticipated H3K9ac increase. mESC MNase-seq was normalized with the background factor sb alone; flanking occupancy fell and the +1 nucleosome rose at promoters. · 3.6 Ryder preserves expected global changes in…; Figure legends

4. SuggestedSpike-in normalization did not act as a fixed gold standard. Read-ratio and profile-based spike-in corrections gave different results, and CTCF-referenced Ryder called more decreased sites than spike-in read ratio in the BRG1 degron DNase-seq.

   In BRG1-AID fibroblasts with human 293T spike-in cells, all methods showed reduced enhancer accessibility, but internal-reference normalization found more significantly decreased enhancer and promoter sites than RPM or spike-in ratio. In MV411 ATAC-seq with a Drosophila S2 spike-in, the ratio and profile spike-in strategies diverged, and S2 and MV411 signal at the displayed loci looked non-equivalent. · 3.4 DNase-seq analysis of BRG1-AID cells with…; 3.5 Ryder recovers BRG1-dependent enhancer…; Figure legends

5. SuggestedIn ATAC-seq of resting versus stimulated CD8+ T cells, peaks at transcription start sites (TSSs) of stably expressed genes, chosen from matched RNA-seq, worked as a reference set in place of CTCF sites. They kept the activation-linked global gain in accessibility.

   The TSS references came from 500 genes with the smallest fold changes and >1 RPKM in both conditions. With either reference set, accessibility rose significantly at TSSs of upregulated genes. At Ifng, RPM showed an apparent post-stimulation drop at the promoter and nearby enhancers, whereas Ryder preserved the gain that matches the gene's induction. · 3.3 Stable TSSs can serve as alternative…