Oncofusion

start / graph / ana-q130-all-three-deposited-inputs-carry-their-own-monotone-eviction-gradient-of-about-1-15-fold

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q130-does-the-eviction-gradient-exist-in-h3k79me2

The four eviction quartiles are not equivalent stretches of genome: all three GSE260742 input tracks give a monotone increasing share across Q1 to Q4 at Spearman -1.000, 1.15 / 1.17 / 1.13-fold - small, real, in the same direction as the H3K79me2 gradient, and it cancels exactly in a within-genotype drug/vehicle ratio

also called the input gradient · the quartiles differ in genomic DNA too · why G5 was recorded uninformative

Even before any antibody is used, the raw DNA from these cells is not evenly represented across the four groups of stretches - there is about fifteen percent more of it in one group than another, in a steady trend. It is small, and it drops out of the comparison being made, but it means the groups are not interchangeable.

The same thing, for a scientist

**Measured on the three deposited input tracks**, binned at 2 kb on the same mm10 grid over the same universe and scored over the same four eviction quartiles [E1]: | input | Q1 | Q2 | Q3 | Q4 | Spearman vs depth | max/min | |---|---:|---:|---:|---:|---:|---:| | NT (GSM8123416) | 0.000750 | 0.000785 | 0.000832 | 0.000863 | **-1.000** | 1.1507 | | PCGF1-KO C1 (GSM8123415) | 0.000771 | 0.000820 | 0.000898 | 0.000901 | **-1.000** | 1.1686 | | timecourse (GSM8123451) | 0.000532 | 0.000547 | 0.000589 | 0.000601 | **-1.000** | 1.1297 | **All three, independently, perfectly monotone in the same direction as the H3K79me2 drug response.** That is a property of the bin sets - copy number, sonication accessibility, mappability, in some combination an input track cannot separate - and it is a reason to distrust any single-track share comparison across these quartiles [E1]. **Why it does not damage this beat's result, stated as arithmetic rather than reassurance.** The scored statistic is `R = (drug share over the quartile) / (that genotype's own DMSO share over the same quartile)`. **The input appears in neither term**, so an input feature common to both cancels exactly. Empirically: input-normalising every value changes the control VTP profile from `0.5860 / 0.8188 / 0.9280 / 1.1865` to `0.5785 / 0.8095 / 0.9209 / 1.1877` - **a maximum change of 0.0093** - and leaves the Spearman at -1.000 [E2]. **And the direction matters.** The input gradient runs the SAME way as the observed one, so if it were biasing anything it would be biasing it towards a *smaller* observed effect; correcting for it makes the gradient marginally steeper, not shallower (Q4 rises from 1.1865 to 1.1877 while Q1 falls from 0.5860 to 0.5785) [E2]. **This is why the pre-registered criterion G5 is recorded UNINFORMATIVE rather than SURVIVES.** The pre-registration said G5 would be uninformative if the input carried a gradient. It does. The clause fires, and the verdict is recorded as the clause demands even though the result survives [E3].

Go deeper
What a 1.15-fold input gradient could be, and why it cannot be resolved here

An input track measures fragmented, cross-linked genomic DNA. A 15% monotone excess at the least-evicted quartile is consistent with at least three things that cannot be separated from a single input: copy number (no copy-number measurement is deposited for any GSE260742 culture), chromatin accessibility to sonication (open chromatin fragments more readily, and the least-evicted sites retain the fusion and presumably its activity), and mappability (the blacklist removes the worst regions but not all of them).

The accessibility reading is the most interesting of the three, because it would mean the quartile split is partly a split on chromatin state and not purely on drug-driven eviction. Nothing in this beat tests that, and it is the most substantive unexamined confound in the whole line of work, because it would apply identically to Q117's, Q125's and this beat's gradients.

Facts and figures
scriptanalyses/q130-does-the-eviction-gradient-exist-in-h3k79me2/run.py
statusrun
result summaryAll three inputs Spearman -1.000 across the eviction quartiles at 1.1507 / 1.1686 / 1.1297 max/min; input normalisation changes the H3K79me2 quartile values by at most 0.0093.
absence classmeasured-not-analysed
absence class noteQ124 binned all three inputs and used them only to define universes and compute floors; Q125 used the matched inputs for a clonal-drift screen and a CpG-island correction. Nobody had asked whether the inputs are graded across the eviction quartiles the whole line of work rests on.
absence class assigned2026-09-23 (Q130)
epigenetic readoutinput (genomic DNA) share across fusion-eviction quartiles, as the structural control for every mark scored on them
input gradients
NT_INPUT-1.000NT_INPUT: -1C1_INPUT-1.000C1_INPUT: -1TC_INPUT-1.000TC_INPUT: -1
inputs
[
  "GSM8123416 NT input",
  "GSM8123415 PCGF1-KO input",
  "GSM8123451 timecourse input"
]
input gradients — the numbers behind the chart
{
  "NT_INPUT": {
    "share_by_quartile": [
      0.00075,
      0.000785,
      0.000832,
      0.000863
    ],
    "spearman_vs_depth": -1,
    "max_over_min": 1.1507
  },
  "C1_INPUT": {
    "share_by_quartile": [
      0.000771,
      0.00082,
      0.000898,
      0.000901
    ],
    "spearman_vs_depth": -1,
    "max_over_min": 1.1686
  },
  "TC_INPUT": {
    "share_by_quartile": [
      0.000532,
      0.000547,
      0.000589,
      0.000601
    ],
    "spearman_vs_depth": -1,
    "max_over_min": 1.1297
  }
}
input normalised NT VTP
[
  0.5785,
  0.8095,
  0.9209,
  1.1877
]
raw NT VTP
[
  0.586,
  0.8188,
  0.928,
  1.1865
]
Still open
  1. Is the 1.15-fold input gradient copy number, accessibility or mappability? No copy-number, ATAC or MNase data is deposited for any GSE260742 culture, so it cannot be resolved from this series at all.
  2. If it is accessibility, does the eviction-quartile split partly measure chromatin state rather than drug-driven eviction? That would apply identically to Q117's, Q125's and this beat's gradients and is the largest unexamined confound in the line of work.
  3. Do Q125's marks show the same relationship to the input gradient? The check is the same script and was not run for them.
Every source behind this entry (3)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    All three GSE260742 input tracks - GSM8123416 (NT), GSM8123415 (PCGF1-KO) and GSM8123451 (timecourse) - carry a monotone increasing within-sample share across the four eviction quartiles of S_FLAG at Spearman -1.000, with max/min ratios of 1.1507, 1.1686 and 1.1297.

    What this evidence cannot show

    NO SPIKE-IN NORMALISATION IS POSSIBLE. Q117 verified zero hits for the entire spike-in vocabulary across all 419,824 characters of GSE260742's 158 sample records and no scale factor is deposited, so every cross-condition number here is a WITHIN-SAMPLE SHARE. It cancels the unknown per-sample scale factor exactly and IS BLIND TO A UNIFORM CHANGE BY CONSTRUCTION: nothing here measures H3K79me2's abundance, everything measures its distribution. EXPOSURE ASYMMETRY: DMSO 72 h and SGC0946 5 uM 72 h are matched, VTP-50469 500 nM is 48 h, so every VTP ratio is a LOWER BOUND on the 72-hour effect - identical in both genotypes, so the genotype DIFFERENCE is unaffected. REPLICATION, CORRECTED MID-BEAT: the NTvKO arm deposits ONE H3K79me2 ChIP per genotype per drug, so ITS spans carry no technical error bar and its bin-level bootstrap intervals measure sampling within one library, NOT library-to-library variation. A REPLICATE PAIR DOES EXIST for a different culture - GSM8123372/374 and GSM8123386/388, both titled MLLAF9 Control, the second carrying `rep2` - already binned by Q122 and scored in Q130 DEVIATION 4: the SIGN and the monotonicity reproduce (Spearman -1.000 in all four murine VTP contrasts in this deposit) while the SPAN reproduces only to within a factor of 1.99 (-0.2438 against -0.4847). EVERY SPAN IN THIS BEAT SHOULD BE READ WITH THAT 2-FOLD REPRODUCIBILITY. THE KNOCKOUT is one CRISPR guide, one clone C1, no second clone, no rescue, and NO deposited PCGF1 protein or transcript measurement. S_FLAG is defined by a DYKDDDDK-tag ChIP on a RETROVIRALLY OVER-EXPRESSED tagged MLLAF9 construct, so it is the over-expressed molecule's sites and not an endogenous fusion's. THE H3K79me2 ANTIBODY IS NOT RESOLVED - `antibod` occurs zero times in the 158 GEO records - and is recorded unresolved rather than guessed. SYSTEM: murine MLL-AF9 (KMT2A::MLLT3) myeloid blood cells, NOT the infant B-lineage KMT2A::AFF1 disease this graph is about. INPUT-SPECIFIC: an input track cannot separate copy number from sonication accessibility from residual mappability, and no copy-number measurement is deposited for any culture in this series. Three inputs agreeing is three tracks from one laboratory on one genome, not three independent estimates.

    Evidence that argues against this (1)
    • Three perfectly monotone inputs could also mean the quartile split is partly a split on chromatin accessibility, which would make the 'eviction depth' label partly a misattribution. Iqbal Lab oncofusion knowledge graph 2026
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    Input-normalising the control VTP contrast changes the four quartile values from 0.5860 / 0.8188 / 0.9280 / 1.1865 to 0.5785 / 0.8095 / 0.9209 / 1.1877 - a maximum change of 0.0093 - and leaves the Spearman at -1.000, and the correction makes the gradient marginally steeper rather than shallower.

    What this evidence cannot show

    The input-normalised statistic divides each bin's share by the input's share and sums, which is a different estimator from the ratio of sums and not simply a corrected version of it; bins with zero input are dropped. It is quoted as a sensitivity check, not as the primary number.

    Evidence that argues against this (1)
    • Agreement between a raw and an input-normalised statistic computed from the same tracks is close to guaranteed when the input gradient is small, so it is a weak check. Iqbal Lab oncofusion knowledge graph 2026
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    Q130's pre-registered criterion G5 declared itself uninformative if the input carried an eviction gradient, the clause fired, and G5 is recorded UNINFORMATIVE BY ITS OWN RULE rather than as SURVIVES.

    What this evidence cannot show

    The clause was written as a bare sign test with no magnitude threshold, which at n = 4 makes it fire on any monotone input however small. That is a defect in the clause and is recorded as one in the deviation.

    Evidence that argues against this (1)
Connected to