start / graph / hyp-q135-the-same-condition-floor-is-a-property-of-the-stratum-definition-not-of-the-replicate-pair

HypothesisspeculativeQ135-the-clean-pair-and-the-two-uncomputed-mark-channels

Hypothesis: the same-condition floor measured this way is largely a property of where the EXTERNAL stratifier puts its top bins relative to each library's own coverage, not of how well the two replicates agree - which would explain a twentyfold arm span sitting on top of rank concordances that barely differ.

also called the floor measures the stratifier's fit · not agreement, alignment · why 20x on top of 0.65-0.88

A puzzle: repeat measurements that agree about equally well produce wildly different reliability numbers. The proposed explanation is that the number depends less on the repeats agreeing with each other than on how well a third, separate dataset used for sorting happens to line up with each of them.

The same thing, for a scientist

**The observation to explain.** Rank concordance across the deposit's same-condition H2AK119ub pairs spans 0.6481-0.8796 with no clean separation by arm, while the D1 floor spans 1.89%-37.04% - a twentyfold range - and the pair with the LOWEST concordance (C1_VTP, 0.6481) gives the TIGHTEST floor (0.40%). Concordance and floor are not tracking each other and may be anti-correlated. **The hypothesis.** `R(S)` is a ratio of share masses inside a stratum `S` defined on a third track (FLAG DMSO, GSM8123359). Its value depends on how much of each library's share mass falls inside `S`. If `S`'s top bins sit in a region where library A is locally deep and library B is locally shallow, `R` departs from 1 even when A and B agree everywhere in rank. **The MLLAF9 Control arm's libraries may simply be less well aligned to a stratifier drawn from a FLAG ChIP of the tagged construct** - which is the arm the stratifier itself comes from, so this is not an obvious direction to predict. **Why it matters.** If true, the floor is not a noise floor at all but a stratifier-fit statistic, and the entire practice of scoring effects against it - cycles 100-108 - is measuring effects against a quantity that partly encodes the stratum definition. That would not invalidate sign or gradient results, which are computed within a fixed stratification, but it would invalidate magnitude thresholds.

Go deeper

Falsifiers - stated to be scoreable

F1. Recompute every pair's floor under a SECOND external stratifier (the already-binned q117_bin_KDM2B_DMSO.tsv or q117_bin_INPUT_B2.tsv, neither a member of any pair). If the arm ranking by D1 floor is preserved - PCGF1-KO tightest, MLLAF9 Control widest, with the span staying above 5x - the hypothesis is REFUTED and the floor is a property of the pairs. If the ranking reorders or the span falls below 2x, it is SUPPORTED. This is zero downloads and one re-run.

F3. Self-stratify one pair (forbidden as an analysis, legitimate as a probe). A sibling beat measured that self-stratification manufactures a monotone decile gradient of span 0.331 from two identical libraries. If the floor under self-stratification is within 1.5x of the external-stratifier floor for the tight arms but not the wide one, the stratifier-fit reading is SUPPORTED.

What would make this hypothesis uninformative

If the alternative stratifiers in F1 are themselves drawn from the same arm and share the FLAG track's coverage structure, F1 cannot discriminate. q117_bin_INPUT_B2.tsv (GSM8549823, an Oct-2024 input) is the better choice precisely because it comes from the other submission batch, and F1 should be run with it as well as with KDM2B.

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.

Facts and figures

status	stated cycle 108, none of F1-F4 run

analyses

[
  "F1: re-score all pairs under q117_bin_KDM2B_DMSO.tsv and q117_bin_INPUT_B2.tsv; zero downloads, one re-run",
  "F2: regress |R-1| at TOP on the ratio of the two members' share-inside-TOP across the nine pairs",
  "F3: self-stratify one pair as a probe and compare to its external-stratifier floor",
  "F4 (decisive, more work): depth-matched downsampling of the deeper member of each pair to the shallower member's bin count, then re-score"
]

falsifiers

[
  "F1 REFUTES if the arm ranking by D1 floor is preserved and the span stays above 5x under a second external stratifier",
  "F2 REFUTES if |Spearman| between |R-1| at TOP and the share-inside-TOP ratio is at or below 0.4 across the nine pairs",
  "F3 SUPPORTS if self-stratification changes the wide arm's floor by more than 1.5x while leaving the tight arms within 1.5x",
  "The hypothesis is WITHDRAWN outright if F1 and F2 both refute"
]

Still open

Is a stratifier drawn from the arm under test a legitimate instrument at all, or does it need to come from outside the arm as well as outside the pair?
If F4 shows depth explains it, does every floor in cycles 100-108 need re-scoring at a common depth?

Every source behind this entry (1)

E1weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.dataset
Across nine GSE260742 same-condition H2AK119ub comparisons, Spearman rank concordance spans 0.6481-0.8796 while the D1 floor spans 0.40%-45.72%, and the least concordant pair (C1_VTP, Spearman 0.6481) yields the tightest floor (D1 0.40%). Concordance does not predict the floor in this sample.
Q135 re-analysis of GSE260742: the same-condition H2AK119ub floor decomposed into processing-marker and submission-batch terms across three experimental arms, plus first per-mark floors for H3K79me2 and H3K27me3. analyses/the-clean-pair-and-the-two-uncomputed-mark-channels/{PREREGISTRATION.md,run.py,results.json}, run 2026-09-23. 21 tracks downloaded one at a time, binned at 2 kb with Q117Bin, each raw .tdf deleted before the next fetch; 2,322,896,326 bytes transferred, 205,959,406 bytes peak raw disk. All 21 accessions re-verified live against the GEO record before use. · Iqbal Lab oncofusion knowledge graph 2026
What this evidence cannot show
n = 9 comparisons and the hypothesis is generated from the same nine, so it is a POST-HOC explanation of an observation and is written as a hypothesis with pre-stated falsifiers rather than as a finding. None of F1-F3 has been run. SYSTEM: murine MLL-AF9 (KMT2A::MLLT3) myeloid blood cells, Mus musculus, mm10, one lab, one deposit - NOT the infant B-lineage KMT2A::AFF1 disease this graph is about. n = 3 conditions per arm, one cell system; the spread across conditions is reported so stability is visible and IS NOT A CONFIDENCE INTERVAL. 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)
- Iqbal Lab oncofusion knowledge graph 2026
- Genome Research 2012

Connected to

related to →The same-condition H2AK119ub noise floor in GSE260742 spans twentyfold across the deposit's three experimental arms - D1 |R-1| of 1.89% in the PCGF1-KO clone, 10.49% in the non-targeting control and 37.04% in the MLLAF9 Control arm - at one mark, one lab, one pairing shape and one external stratifier. A single same-condition floor for this deposit does not exist.
the candidate mechanism for the twentyfold arm span: if it holds, the span is a property of the stratum definition rather than of the replicate pairs
depends on →Across every same-condition H2AK119ub replicate pair in GSE260742 - nine pairs, three arms, both pairing shapes - rank concordance over the shared bin universe runs 0.6481 to 0.8796 and never reaches 0.88. The floors this thread scores against are all computed from pairs that are only moderately concordant.
the concordance range that makes the span puzzling in the first place
depends on →Every noise floor in this thread is computed from within-sample shares, which sum to one. A purely local change therefore leaks into every other stratum by arithmetic, so no floor measured this way can distinguish 'the bulk moved' from 'the peaks moved and the shares renormalised'. A tighter floor is not thereby a cleaner one.
the share construction that makes the floor sensitive to where the stratum sits
← testsRe-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.
the pre-registered F1 falsifier, promoted to a runnable experiment