The primary donor-aware analysis detected zero regions meeting the declared
FDR and effect-size thresholds. This is the controlling result.
The dataset contains four complete donor pairs and two Memory-only donors.
The model adjusts for donor, but the design remains small and unbalanced.
Peak-universe construction used condition-level unions and retained peaks
supported by a single sample. Conclusions depend on this declared universe.
Count-assignment rates varied substantially between samples and are technical
context, not a basis for post hoc exclusion.
Raw-p-value candidate sets are nominal. They do not control the false
discovery rate.
Apeglm shrinkage reduced many ordinary effects, emphasizing uncertainty in
low-information estimates.
Nearest-gene labels do not demonstrate enhancer-gene regulation or gene-
expression changes.
g:Profiler terms depend on nominal region selection and nearest-gene mapping.
They do not prove pathway activation.
A HOMER motif label describes sequence similarity and enrichment relative to
a background. It does not demonstrate transcription-factor expression,
binding, occupancy, or causality.
IGV tracks were independently autoscaled and were used for qualitative visual
QC, not normalized group comparison.
Week 1 used one Immature NK sample and a chromosome-22 teaching subset. Its QC
cannot be generalized to the full ten-sample comparison.
No footprinting analysis was completed; the repository makes no footprinting
or differential-binding claim.