starplast.chromatin

Chromatin signal per gene, summarised from coverage tracks over promoters.

GEO serves these two experiments as bigWig only -- no peak calls, no per-gene table -- so the number has to be computed here. What is computed is deliberately the dullest possible summary: the mean coverage over a fixed window around each gene's transcription start, normalised to the genome mean. No peak calling, no background model, no thresholds. Every one of those is a modelling choice that would be invented here rather than taken from the authors, and the project's rule is that a number which looks like a measurement must be one.

Why the promoter and not the gene body

Accessibility and factor occupancy are promoter phenomena; over a gene body they mostly report how long the gene is. The window is PROMOTER bases either side of the start, taken from the strand, so a reverse-strand gene's promoter is at its higher coordinate. Getting that backwards is silent -- the column still looks like data -- which is why the loader is tested on a reverse-strand gene.

What only the unperturbed arm is used

Both deposits are knockdowns, and both include an untreated arm. Only the untreated arm is read: it is the measurement of what chromatin looks like in a normal parasite, which is what the slot asks. The perturbed arms measure what a specific depletion does, which is a different question and would need its own slot rather than being averaged into this one.