Oncofusion

start / graph / ana-paired-is-a-replicate-label-and-gse260742-has-seventeen-same-condition-replicate-groups

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.

The deposit names its first replicate "paired" and its later ones rep1/rep2/rep3, so GSM8123403 "H2AK119ub paired" and GSM8123420 "H2AK119ub rep2" are one condition in two replicates - which this project has called "the two processing conventions" since cycle 88. GSE260742 holds 17 genuine same-condition replicate groups, 13 of them histone marks, refuting the claim repeated since cycle 100 that it has none and never will

also called paired is a replicate label · 17 same-condition replicate groups · the conventions were replicates all along · GSE260742 does have a selection control

A public dataset labels repeated runs of the same experiment in an inconsistent way: the first repeat is called "paired" and later ones "rep2" or "rep3". Because of that, this project spent nineteen cycles treating two repeats of the same experiment as two different ways of processing one experiment, and separately concluded that the dataset contained no repeats at all. Reading the dataset's own list of 158 samples shows seventeen sets of repeats, thirteen of them for the chemical marks the project measures.

The same thing, for a scientist

**Statement.** Parsing all **158 sample records of GSE260742** live from GEO settles three questions at once, and overturns a premise this project has carried since cycle 88. **1. `paired` IS A REPLICATE LABEL, NOT A PROCESSING CONVENTION.** The deposit titles its replicates `paired`, `rep1`, `rep2`, `rep3`. Verbatim: | accession | title | depositor's filename | |---|---|---| | `GSM8123403` | NTvKO Control - DMSO H2AK119ub **paired** | `ChIP_K119UB_F3_NT_DMSO.filt.bam.tdf` | | `GSM8123420` | NTvKO Control - DMSO H2AK119ub **rep2** | `ChIP_K119UB_NT_DMSO.tdf` | **Same cell system, same genotype, same drug, same mark - two different GEO samples.** Q125 maps exactly these to `K119_PAIRED_NT_DMSO` and `K119_REP2_NT_DMSO`. **Cycles 88, 104, 105 and 106 called that pair "the two processing conventions" and measured how often they disagree. They are two biological replicates that also differ in processing state** - confounded, which is what Q125's own `G2` caveat said before the replicate half was dropped. **2. THE DEPOSIT HAS 17 GENUINE SAME-CONDITION REPLICATE GROUPS**, 13 of them histone marks: H2AK119ub at **three** replicates in each of three MLLAF9 conditions, H3K27me3 and H3K79me2 at two each, RING1B at two, plus the six NTvKO H2AK119ub `paired`/`rep2` pairs. **The claim repeated from cycle 100 to cycle 106 - that GSE260742 has no same-condition replicate and never will - is false**, and it sent a 1.87 GB download after a control that was already on disk. **3. THE SIX FLAG MISLABELS REPRODUCE EXACTLY.** Comparing every title against the depositor's filename flags **6 conflicts and only 6** - `GSM8123407-412`, titled *H2AK119ub rep1*, filenames `ChIP_FLAG_*`. Cycle 101 found these by hand; an independent instrument recovers the same six, no more and no fewer. **4. H3K79me2's titles are self-consistent** - all 30 samples' filenames name the same antibody - so cycle 106's "replicates" reading stands, now checked rather than inferred. **And H3K27me3 does have a second channel nobody computed** (`GSM8549820/21/22`), upgrading cycle 106's weak negative to `C2`.

Go deeper
Two of my own criteria were wrong, and both returned the comfortable answer

The suffix criterion. I first asked whether a mark declares two distinct rep suffixes. But a deposit that names its first replicate by saying nothing has two channels carrying one suffix - X and X rep2. That returned C1, "H3K27me3 has no second channel", when three exist. Counting conditions rather than suffixes inverted it to C2.

The group criterion. I then discarded any replicate group containing a FLAG mislabel. That threw away the genuine pair sitting inside it: NTvKO Control - DMSO H2AK119ub holds the mislabelled GSM8123410 and the real pair GSM8123403/GSM8123420 - exactly the pair Q125 used. Dropping the bad sample rather than the group moved the count from 10 to 17.

Both errors, and the rep-only regex that started this, failed in the same direction: fewer replicates, less to answer for. That is three filters in two cycles that quietly removed the finding.

Facts and figures
questionwhat does GSE260742 actually say about its own channels and replicates?
samples parsed158
fetchedlive from NCBI GEO 2026-09-23
downloads of data0
genuine same condition replicate groups17
of which histone marks13
title filename conflicts6
refutesthe claim repeated from cycle 100 to cycle 106 that GSE260742 has no same-condition replicate and never will
absence classmeasured-not-curated
bite backThis entry says a control exists where five cycles said none did. If any of the 17 groups turns out to be confounded in a way Q131 already identified, the practical conclusion collapses back to Q131's and the only thing left standing is the naming point.
replicate labels used by the deposit
[
  "paired",
  "rep1",
  "rep2",
  "rep3"
]
the pair that was called two conventions
{
  "GSM8123403": "NTvKO Control - DMSO H2AK119ub paired -> tag K119_PAIRED_NT_DMSO",
  "GSM8123420": "NTvKO Control - DMSO H2AK119ub rep2 -> tag K119_REP2_NT_DMSO",
  "reading": "one condition, two replicates, also differing in processing state"
}
the six
[
  "GSM8123407",
  "GSM8123408",
  "GSM8123409",
  "GSM8123410",
  "GSM8123411",
  "GSM8123412"
]
h3k27me3 second channel
[
  "GSM8549820",
  "GSM8549821",
  "GSM8549822"
]
h2ak119ub third replicate
[
  "GSM8549830",
  "GSM8549831",
  "GSM8549832"
]
my own defects
[
  "a rep<N>-only regex that could not see the label 'paired'",
  "a criterion requiring two DISTINCT suffixes, which returned C1 wrongly",
  "discarding whole replicate groups containing one mislabelled member, which lost the very pair Q125 used"
]
Still open
  1. Re-read Q131's vehicle-pair adjudication against these 17 groups explicitly. It concluded every vehicle pair is confounded; this entry shows replicates exist but does not show any is clean, and the two must be reconciled rather than left side by side.
  2. Re-title every published statement about 'the two H2AK119ub conventions' to say 'two replicates, confounded with processing state'. Cycles 88, 104, 105 and 106 all carry the wrong framing.
  3. Score the H3K27me3 second channel (GSM8549820/21/22) and the H2AK119ub third replicate (GSM8549830/31/32). Neither has ever been computed.
  4. Do the other deposits this project uses label replicates inconsistently too? One regex missed 'paired' here; GSE127507 and GSE299100 were never checked for the same trap.
Every source behind this entry (1)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.analysis

    All 158 sample records of GSE260742, fetched live from GEO, show the deposit labels replicates paired/rep1/rep2/rep3. GSM8123403 is titled 'NTvKO Control - DMSO H2AK119ub paired' with supplementary file ChIP_K119UB_F3_NT_DMSO.filt.bam.tdf and GSM8123420 'NTvKO Control - DMSO H2AK119ub rep2' with ChIP_K119UB_NT_DMSO.tdf; Q125's run.py maps these to K119_PAIRED_NT_DMSO and K119_REP2_NT_DMSO. Stripping the replicate label and grouping by the remaining title gives 17 genuine same-condition replicate groups after dropping the six mislabelled samples, 13 of them histone marks. Comparing each title's antibody against its filename's antibody flags exactly six conflicts, GSM8123407 through GSM8123412, all titled H2AK119ub rep1 and all named ChIP_FLAG_*.

    What this evidence cannot show

    A GEO TITLE AND A FILENAME ARE BOTH WRITTEN BY HUMANS. Where they agree this shows the depositor was self-consistent, not that the antibody was correct - the six conflicts prove the failure mode exists in this very deposit. 24 samples carry filenames naming no antibody and are reported as UNDETERMINED rather than dropped silently. Grouping is by exact title after stripping the replicate label, so two samples of one condition titled inconsistently would not group, making 17 a LOWER bound. That paired/rep2 are biological replicates rather than two sequencing runs of one library is the deposit's claim, not something this analysis can verify; either way they are two samples, not two processings of one sample. The processing difference (.filt.bam.tdf vs bare .tdf) is real and means the pair is confounded, exactly as Q125's G2 said.

    Evidence that argues against this (3)
    • The strongest thing said here - that 19 cycles misread a replicate pair as a processing convention - is partly unfair to those cycles. Q125's G2 caveat states the confound explicitly: 'the conventions are confounded with processing state (paired is .filt.bam.tdf, rep1 and rep2 are bare .tdf)'. The information was on the record from the start. What went wrong is narrower and less dramatic: later beats quoted the processing half of that sentence and dropped the replicate half.
    • '17 genuine same-condition replicate groups' counts groups, not usable controls. Q131 examined vehicle pairs in this deposit and concluded every one was confounded; this entry does not re-open that adjudication and does not show any of the 17 is clean. The honest claim is that same-condition replicates EXIST, which contradicts 'none and never will', not that a clean selection control exists.
    • The six-conflict reproduction of cycle 101 is weaker evidence than it looks. My instrument searches filenames for antibody tokens, which is the same method cycle 101 used; agreement therefore shows the method is repeatable, not that it is right. An antibody mislabelled consistently in BOTH title and filename is invisible to both of us.
Connected to