The spike-in ruler this thread has been blocked on for eight cycles is deposited as a 2 MB text file, and it includes RS4;11 - the index fusion: under DOT1L inhibition the fly fraction rises 4.9x to 23.4x against matched DMSO, so H3K79me2 collapses GLOBALLY by roughly that factor, and a within-sample share would have shown almost none of it
also called the spike-in ruler exists · 4.9x to 23.4x · the compositional failure, measured · GSE149183 deposits dm6 counts and ratios
All of this project's chromatin measurements are proportions - what share of a signal sits in one place versus another. Proportions cannot detect a change that affects everything at once, because the shares still add to one. Fixing that needs a known amount of fruit-fly material added to each sample as an external ruler, and the project had concluded no such ruler was available. One is, in a public two-megabyte text file, for a cell line carrying the exact fusion this project studies. It shows that a drug of the kind used here can wipe out the signal across the whole genome by five- to twenty-fold - a change the proportion-based method would barely register.
The same thing, for a scientist
**Statement.** Two beats yesterday and one earlier today converged on the same blocker: within-sample shares sum to 1, so they cannot see a global change, and no spike-in was available. **A single GEO query on the submitting author plus `spike-in` returns `GSE149183`**, whose processing line is the same lab's boilerplate with the opposite outcome - *"Global differences in fly-human chromatin ratios **were found**, so H3K79me2 TSS signals were normalized accordingly"* - and whose series-level supplementary files carry **the dm6 counts and the per-sample ratios, in 1.9 MB of plain text.** **One of the two cell lines is `RS4;11` - a KMT2A::AFF1 line. Almost every chromatin result in this thread is KMT2A::MLLT3.** **RS4;11, H3K79me2, depositors' own dm6 counts:** | day | DMSO % fly | EPZ5676 1 µM | compound 10 100 nM | |---|---:|---:|---:| | d3 | 0.933 | 8.687 (**9.31×**) | 4.583 (**4.91×**) | | d6 | 1.320 | 8.184 (**6.20×**) | 23.345 (**17.68×**) | | d9 | 1.769 | 17.533 (**9.91×**) | 34.091 (**19.27×**) | MOLM13 runs in parallel and reaches **23.41×** at d9. **`R1` fires, and not marginally.** The spike-in is a fixed input, so a fly fraction rising **4.9× to 23.4×** means the human H3K79me2 signal fell **globally** by about that factor. **Every noise floor this project has measured is 2-16%. These are changes of 400-2300%.** A within-sample share renormalises to 1 and **would have shown almost none of it** - which is the compositional failure mode this thread has been warning about, now measured rather than feared. **`M2`, and it is the disappointment.** The spike-in table is **DOT1L-inhibitor only** - DMSO, EPZ5676, compound 10. GSE149183 *does* hold MEN1 and DOT1L ChIPs under **VTP-50469**, but they are not in the spike-in-normalised H3K79me2 series. **So the menin inhibitor cannot be put on this ruler from the deposited tables, and cycle 83's standing threat - a DOT1L inhibitor moving the MLL1 channel as much as a menin inhibitor - stays untested on an absolute scale.** What changed is that the experiment which would test it is now exactly specified.
What this does and does not do to the existing results
It does not overturn them. These are TSS-averaged sitepro values over a 1 kb-upstream / 3 kb-downstream window, in a different deposit, for H3K79me2 under a DOT1L inhibitor. Cycles 102-107 score binned genome-wide shares, mostly for H2AK119ub under a menin inhibitor. Nothing here is substitutable into those analyses and U4 says so.
What it does is price the caveat. Until now "shares cannot see a global change" was an argument. It is now a measurement with a magnitude attached, in the right disease and - for the first time in this thread - in the right fusion. Any future share-based claim has to say why a 5-23× global effect could not be hiding underneath it.
And it sharpens one specific worry. Cycle 107's beat withdrew "the chromatin effect is confined to the peaks" for the marks, and named the deflationary alternative: a purely peak-localised gain produces a bulk profile by arithmetic. This entry shows the arithmetic is not hypothetical - a drug in this class really can move the whole distribution far enough to dominate any share.
Two traps in the files themselves
The two supplementary files are not the same format. RS4_11 is tab-separated and MOLM13 is comma-separated, and their spike-in headers differ - hg19/mm9 PF_READS_ALIGNED against hg19 PF_READS_ALIGNED. Parsing both with one delimiter yields a single column and no error.
And the spike-in block is not a separate file. It sits in columns 13-17 of a per-gene table, beginning three rows down, with a blank spacer column before it. A reader that stops at the gene columns never sees it - which is plausibly why the corpus spent eight cycles believing no ruler existed while this file was public.
| question | does a deposited spike-in ruler exist, and how large is the global change it reveals? |
|---|---|
| series | GSE149183 |
| subseries of | GSE149185 |
| bioproject | PRJNA627612 |
| found by | one GEO query: submitting author AND spike-in |
| file size MB | 1.9 |
| downloads | 2 files, 3.9 MB total |
| RS4 11 is | a KMT2A::AFF1 line - this project's index fusion, and a first for its chromatin thread |
| verdict R1 | the compositional worry is real and large - 400 to 2300 per cent global change |
| verdict M2 | no VTP-50469 arm in the spike-in table, so the menin inhibitor cannot be put on this ruler |
| what it does not do | TSS-averaged sitepro in a different deposit for a different mark under a different drug class - NOT substitutable into cycles 102-107 |
| bite back | This entry argues that a class of drug can abolish a histone mark genome-wide, on the strength of the depositors' own alignment counts. If their dm6 alignment was permissive, the fly counts and therefore the fold-changes are inflated, and this entry will have been used to discount share-based results that were fine. The numbers are reported verbatim and were not re-derived from reads. |
| evidence state | curated-not-measured |
| evidence state why not absence class | This entry is not an absence. It reached for absence_class to record an evidence state that vocabulary cannot express - 'curated-not-measured' - so the value moved to evidence_state rather than being folded into an absence token that would misdescribe it. See q190: only 69 of the 529 entries carrying absence_class were absence entries. |
rs4 11 PCT fly — the numbers behind the chart
{
"DMSO_d3": 0.933,
"EPZ5676_d3": 8.687,
"comp10_d3": 4.583,
"DMSO_d6": 1.32,
"EPZ5676_d6": 8.184,
"comp10_d6": 23.345,
"DMSO_d9": 1.769,
"EPZ5676_d9": 17.533,
"comp10_d9": 34.091
}rs4 11 ratio to dmso — the numbers behind the chart
{
"EPZ5676_d3": 9.31,
"comp10_d3": 4.91,
"EPZ5676_d6": 6.2,
"comp10_d6": 17.68,
"EPZ5676_d9": 9.91,
"comp10_d9": 19.27
}molm13 ratio range
[ 5.41, 23.41 ]
all treatment ratios range
[ 4.91, 23.41 ]
measured noise floors for comparison PCT
{
"TOP": 15.97,
"D1": 3.03
}- Compute the SAME contrast both ways from this one file - within-sample share and spike-in normalised - and show the difference directly. The per-gene columns are right there and the claim that shares would miss this is currently an inference from shares summing to 1.
- Put the menin inhibitor on this ruler. GSE149183 has MEN1 and DOT1L ChIPs under VTP-50469 but they are not in the spike-in series; the fly counts for those libraries would settle cycle 83's standing threat on an absolute scale.
- Read the publication behind GSE149183. This beat used the deposit only, and the paper may report the fly/human ratios GSE127507 never did.
- RS4;11 is the index fusion and this deposit is its first spike-in-normalised chromatin data in this project. The deferred RS4;11 arm should be re-examined against it.
- Re-run the earlier Q118 question against this template: does GSE127507's sibling relationship to GSE149183 mean its 181116 batch is deposited under GSE149185 or another subseries of it?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
GSE149183 deposits two series-level supplementary files containing dm6 spike-in counts and ratios alongside TSS-averaged H3K79me2 signal, for RS4;11 and MOLM13. In RS4;11, human/dm6 aligned reads and fly percentages are: DMSO d3 33391935/314449 = 0.933%, EPZ5676 1uM d3 26737687/2543748 = 8.687% (ratio to DMSO 9.31), compound 10 100nM d3 28632510/1375293 = 4.583% (4.91), DMSO d6 1.320%, EPZ d6 8.184% (6.20), comp.10 d6 23.345% (17.68), DMSO d9 1.769%, EPZ d9 17.533% (9.91), comp.10 d9 34.091% (19.27). MOLM13 spans 5.41 to 23.41 fold. The spike-in table contains no VTP-50469 arm. The processing line states that fly-human ratio differences were found and that H3K79me2 TSS signals were normalized accordingly.
GEO GSE149183, 'Novel Inhibitors of the Histone-Methyltransferase DOT1L Show Potent Antileukemic Activity in Patient-derived Xenografts (ChIP-seq dataset)', SubSeries of GSE149185, BioProject PRJNA627612; series supplementary files GSE149183_RS4_11_ and GSE149183_MOLM13_H3K79me2_TSS-ChIPseq_signal_+_dm6_spike-in_1000u_3000d_res100_sitepro.txt.gz, downloaded and parsed 2026-09-23; reproduced by analyses/the-spike-in-ruler-exists-and-it-is-deposited/run.py · GEO depositors; parsed by the Oncofusion knowledge graph coordinator · NCBI Gene Expression Omnibus 2026What this evidence cannot show
THESE ARE THE DEPOSITORS' OWN NUMBERS AND NOTHING HERE RE-DERIVES THEM FROM READS. If their dm6 alignment was permissive, human reads with fly-like sequence inflate the fly count, and at a 1% baseline that bias is not negligible. A rising fly FRACTION is a ratio of two counts and sequencing depth differs between runs, which is why human counts are reported beside it - note the human count also falls, 33.4M to 13.4M at RS4;11 d9, so the two agree in direction. H3K79me2 is the mark DOT1L writes, so its collapse under a DOT1L inhibitor is the EXPECTED result; the finding is the magnitude and its implication for share statistics, not the direction. These are TSS-averaged sitepro values over -1kb/+3kb, not the binned genome scored in cycles 102-107, and are NOT substitutable into them. No VTP-50469 arm is present, so nothing here bears directly on the menin inhibitor. The deposit is associated with a publication that has not been read in this beat.
Evidence that argues against this (4)
- The headline number is chosen generously. 23.4x is the maximum across eighteen samples and nine of those are DMSO controls with no ratio at all; the minimum treatment ratio is 4.91x and the d3 timepoint - the one most comparable to the 3-4 day treatments used elsewhere in this thread - runs 4.91-9.31x. Quoting '5 to 23 fold' in the title is accurate and also flattering to the point being made.
- The argument that a within-sample share would miss this is asserted, not demonstrated. It follows from shares summing to 1, but the actual per-gene sitepro columns are in the same file and the honest test is to compute share-based and spike-in-normalised versions of the SAME contrast and compare them directly. That was not done here and it is one afternoon of work.
- This deposit was found by a single GEO query that could have been run at any point in the last eight cycles, and the earlier beat in this very cycle published a cost estimate for recomputing from 23 GB of SRA before the query was tried. The standing rule - before publishing a negative, search for the better reagent - was violated within the same cycle in which it was being applied to someone else's work.
- A global collapse of H3K79me2 under a DOT1L inhibitor does not establish that a MENIN inhibitor produces anything comparable, and the absence of a VTP arm here means this entry cannot speak to the drug this project actually studies. Read carelessly it licenses a general scepticism about share statistics that the evidence supports only for a drug class known to abolish its target mark outright.
supersedes that entry's cost conclusion in the same cycle: a sibling deposit from the same lab has the dm6 counts and ratios in 2 MB of plain text, not 23.2 GB of SRA
the same lab's boilerplate with the opposite outcome - here the ratios were found, used, AND deposited
prices that floor's binding caveat: a 5-23x global change is 30-800x the 2-16% floors, and a within-sample share would not see it
the DOT1L-inhibitor half of that experiment is already public in RS4;11 and MOLM13; what is missing is exactly the menin-inhibitor arm
collapses globally 4.9-23.4x under DOT1L inhibition on a spike-in scale, in RS4;11 and MOLM13
discharges that entry's own counter-evidence, which said the blindness of shares was asserted rather than demonstrated - and finds the failure is distortion, not blindness
same deposit family; this survey found GSE173599, a sibling with a degron arm and an EPZ+VTP combination arm that nothing here has opened
the ruler that entry found is deposited for the H3K79me2 samples only; the factor ChIPs in the same series are not in the table
the dm6 ratios that entry reports are unaffected, but its statement that the per-gene RS4;11 columns are usable is wrong: 66.8% of them are MOLM13
its dm6-ratio headline stands; its claim that the deposited per-gene columns are already scaled does not
deposit-specific instance: GSE173599 uses a 21-gene housekeeping panel whose own raw ratios are 0.809 (3 h) and 0.654 (24 h)
- ← related toSpike-in-normalised menin CUT&RUN in a KMT2A::AFF1 degron line, with replicates and a PSIP1 arm
the spike-in requirement in this design is the corpus-wide blocker that entry identified