Inconsistent Signal in CoveragePlot Compared to DE and DA Results for Mki67

[Jyng321]

Hi all,
Thank you, Seurat Team, for an amazing tool.
Currently, I am working with a "low" to "very low" expressing multiomic data.
I am using CoveragePlot() to visualize gene-level activity across two groups: AMC (active - "low") and QMC (quiescent - "very low"). While analyzing the Mki67 locus, I encountered a discrepancy between what CoveragePlot() shows and the results from my differential analysis.
For example, my FindMarkers() output for RNA shows Mki67 as strongly upregulated in the AMC group:
gene = Mki67
avg_log2FC = 1.53
pct.1 = 0.332
pct.2 = 0.023
adjusted p-val = 1.34e-118

And for ATAC peaks near the Mki67 gene, I observe even stronger differential accessibility:
region = chr7-135716118-135716997
avg_log2FC = 3.91
pct.1 = 0.223
pct.2 = 0.013
adjusted p-val = 1.08e-82

The number of cells containing Mki67-fragment is 22.26% in AMC and only 1.28% in QMC.

However, when I use CoveragePlot() with both ATAC and RNA assays for Mki67, the coverage tracks between AMC and QMC appear very similar. In some cases, QMC even shows higher peaks visually, which contradict both the RNA and ATAC DE results.

Moreover, I exported the coverage signal to BED files both with and without scaling and found that the scaled version aligned better with the differential data, suggesting that some internal scaling or normalization step might be affecting the visual interpretation. This raises a couple of questions:

1. Is the signal in CoveragePlot() normalized across cells or groups in a way that could reflect true differences?
2. In this case, which scaling strategy should I use, scaling with CPM, scaling with cell number,...? I want to make sure that I am not biased in showing the peaks.

The detailed information and script are in the attachment.

Thank you for your time and for developing such a powerful toolkit.
Best regards,
Duy.

CoveragePlot_Question-2025-04-30.pdf