Re-score all nine GSE260742 same-condition floors under two further external stratifiers already binned on disk - KDM2B DMSO and the Oct-2024 input - to decide whether the twentyfold arm span is a property of the replicate pairs or of the stratum definition. Zero downloads, one re-run of an existing script.
also called the zero-download follow-up · F1 as an experiment · second stratifier test
The cheapest way to check the surprising result in this analysis is to redo it using a different dataset to sort the genome. Everything needed is already on the machine and it takes one run.
The same thing, for a scientist
**Inputs, all already binned - no download:** | role | file | accession | why | |---|---|---|---| | stratifier 1 (used) | `q117_bin_FLAG_DMSO.tsv` | GSM8123359 | the incumbent | | stratifier 2 | `q117_bin_KDM2B_DMSO.tsv` | GSM8123392 | a different factor, same Mar-2024 batch | | stratifier 3 | `q117_bin_INPUT_B2.tsv` | GSM8549823 | an INPUT, and from the **Oct-2024** batch, so it shares a batch with rep3 and not with rep1/rep2 | | pairs | `C:/Temp/q135/q135_bin_*.tsv`, `C:/Temp/q125/q125_bin_K119_*.tsv` | 21 + 12 | already on disk | **Procedure.** Run `run.py` three times with `STRAT` switched, holding everything else fixed - the same amended universe gate, the same statistic, the same conditions. Report the three arm rankings and the three D1 spans side by side. **Pre-registered reading, so the result cannot be chosen after the fact.** - **The arm ranking is preserved and the span stays above 5x under all three** -> the floor is a property of the replicate pairs. Hypothesis `hyp-q135-...stratum-definition...` is **REFUTED**, and the twentyfold arm span becomes a robust finding that every floor in cycles 100-108 must be re-scored against. - **The ranking reorders, or the span falls below 2x under either new stratifier** -> the floor is substantially a stratifier-fit statistic. Hypothesis **SUPPORTED**, and every magnitude threshold in this thread is scored against a quantity that partly encodes the stratum definition. - **The two new stratifiers disagree with each other** -> the instrument is not stable and no floor from this deposit should be quoted without naming its stratifier. This is the outcome nobody wants and it must be reported if it occurs. **Cost: one re-run, minutes, zero bytes transferred, zero disk.**
Why stratifier 3 is the important one
q117_bin_INPUT_B2.tsv is GSM8549823, an input from the Oct-2024 submission batch. Every floor measured so far is stratified on a Mar-2024 FLAG track, so the incumbent stratifier shares a batch with rep1 and rep2 and not with rep3 - which is precisely the asymmetry the P_batch pairing is supposed to be testing. Re-scoring P_batch under an Oct-2024 stratifier swaps which member of the pair is batch-matched to the stratum definition, and if the floor moves substantially, that alone establishes the stratifier-fit effect without needing F2 or F4.
It is also an INPUT rather than a factor ChIP, so its top bins are accessibility and copy number rather than fusion occupancy - a genuinely different stratification and not a relabelling of the same one.
| cost | zero downloads, one re-run, minutes |
|---|---|
| readout | arm ranking by D1 floor, and the D1 span, under each stratifier |
| status | designed cycle 108, not run |
stratifiers
[
{
"file": "q117_bin_FLAG_DMSO.tsv",
"gsm": "GSM8123359",
"batch": "2024-03-03",
"role": "incumbent"
},
{
"file": "q117_bin_KDM2B_DMSO.tsv",
"gsm": "GSM8123392",
"batch": "2024-03-03",
"role": "different factor"
},
{
"file": "q117_bin_INPUT_B2.tsv",
"gsm": "GSM8549823",
"batch": "2024-10-01",
"role": "input, and batch-matched to rep3 instead of rep1/rep2"
}
]decision rule
{
"ranking preserved and span > 5x under all three": "hypothesis REFUTED; the arm span is a robust property of the pairs",
"ranking reorders or span < 2x under either new stratifier": "hypothesis SUPPORTED; the floor is substantially a stratifier-fit statistic",
"the two new stratifiers disagree with each other": "the instrument is unstable; no floor from this deposit may be quoted without naming its stratifier"
}- Should an annotation-based stratification (CpG islands, already on disk as q117_cgi.json) be added as a fourth arm, since all three track-based stratifiers share a cell system?
- If the three stratifiers disagree, what is the right way to report any floor at all?
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
The three candidate stratifiers and all 33 pair tracks required for this experiment are already binned at 2 kb on local disk: q117_bin_FLAG_DMSO.tsv (GSM8123359), q117_bin_KDM2B_DMSO.tsv (GSM8123392), q117_bin_INPUT_B2.tsv (GSM8549823), C:/Temp/q135/q135_bin_*.tsv (21 files, this beat) and C:/Temp/q125/q125_bin_K119_*.tsv (cycle 125). The experiment requires zero downloads and one re-run of an existing script with one constant changed.
What this evidence cannot show
Scratch directories in this project have been deleted between cycles - `C:/Temp/q118` is gone and took the human binner with it - so 'already on disk' is a claim with a short half-life. If the directories are missing the experiment costs 2.16 GB of re-download for this beat's tracks plus whatever cycle 125's cost. THE RENORMALISATION BOUND, AND IT APPLIES TO EVERY NUMBER IN THIS ENTRY. Within-sample shares sum to 1 and GSE260742 has NO USABLE SPIKE-IN (Q117 verified zero hits for the entire spike-in vocabulary across all 419,824 characters of its 158 sample records, and no scale factor is deposited), so a purely local change leaks into every other stratum by arithmetic. A FLOOR COMPUTED THIS WAY CANNOT DISTINGUISH 'the bulk moved' from 'the peaks moved and the shares renormalised'. A TIGHTER NUMBER IS NOT THEREBY A CLEANER ONE.
Evidence that argues against this (2)
the pre-registered F1 falsifier, promoted to a runnable experiment
whether the twentyfold arm span survives a change of stratifier