GSM8123377 and GSM8123403 carry the same supplementary filename but for the `.filt.bam` marker, yet GEO titles one 'MLLAF9 Control - DMSO H2AK119ub rep2' and the other 'NTvKO Control - DMSO H2AK119ub paired'. Scored directly, they are DIFFERENT libraries (Spearman 0.83), so the collision is a naming accident and both arm titles stand.
also called the filename collision · C-IDENT returns C2 · two files, one name, two libraries
Two samples in a public dataset have essentially the same filename but are labelled as belonging to two different experiments. Comparing them directly shows they really are two different samples, so the labels can be trusted and the matching names are just a slip.
The same thing, for a scientist
**The collision, across all three conditions:** | condition | titled *MLLAF9 Control ... rep2* | titled *NTvKO Control ... paired* | |---|---|---| | DMSO | GSM8123377 `_ChIP_K119UB_F3_NT_DMSO.tdf` | GSM8123403 `_ChIP_K119UB_F3_NT_DMSO.filt.bam.tdf` | | SGC | GSM8123378 `_ChIP_K119UB_F3_NT_SGC.tdf` | GSM8123405 `_CHIP_K119UB_F3_NT_SGC.filt.bam.tdf` | | VTP | GSM8123379 `_ChIP_K119UB_F3_NT_VTP.tdf` | GSM8123406 `_ChIP_K119UB_F3_NT_VTP.filt.bam.tdf` | Identical but for `.filt.bam` - **and both filenames carry `NT`, the NTvKO arm's naming, while one title says MLLAF9 Control.** At most one title per row can be right about the arm, so this had to be settled before the Task-1 pair could be called within-arm. **Pre-registered decision rule and result.** C1 (one library deposited twice) required Spearman > 0.98 **and** every stratum within 1% of unity. Measured: Spearman **0.8436 / 0.8412 / 0.8259** over 1.13-1.17 M shared bins, largest `|R-1|` **21.97%**. **C2 fires - two libraries.** **Calibration matters here and is available.** The same-condition replicate pairs scored elsewhere in this beat give Spearman **0.6481-0.8796**, so 0.83 is squarely inside the range two genuine different libraries produce in this deposit, and nowhere near the identity threshold. The collision is a depositor naming accident.
Why the control was worth running even though it came back negative
Had it returned C1, two things would have followed at once: the .filt.bam marker's isolated effect would have been measurable at zero replicate variance - the cleanest possible measurement of the exact confound cycle 107 flagged - and Task 1's main pairing would have been a cross-arm comparison mislabelled as within-arm, which would have invalidated the beat's headline. A control whose negative result is what licenses the main analysis is worth its cost.
It also leaves a live hazard on the record: this deposit reuses filenames across arms. Any future beat selecting GSE260742 tracks by filename rather than by accession can silently pull the wrong arm. Select by accession.
What none of this settles
It does not establish that the titles are correct - only that the two samples are not the same library, which removes the strongest reason to doubt them. The NT token inside GSM8123377's filename still sits oddly against its MLLAF9 Control title and is recorded unresolved rather than explained. Cycle 101 is this project's standing proof that a GEO title can be wrong, and Q125's DEVIATION 2 found six samples in this very deposit whose titles say H2AK119ub while their filenames and their TDF's own internal track names both say FLAG.
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.
| control | C-IDENT |
|---|---|
| verdict | C2 - two libraries |
| largest stratum floor PCT | 21.97 |
| hazard | GSE260742 reuses supplementary filenames across experimental arms; select tracks by accession, never by filename |
spearman — the numbers behind the chart
{
"DMSO": 0.8436,
"SGC": 0.8412,
"VTP": 0.8259
}replicate pair spearman range same deposit
[ 0.6481, 0.8796 ]
preregistered c1 thresholds
{
"spearman": 0.98,
"max_floor_pct": 1
}- Why does GSM8123377's filename contain `NT` while its title says MLLAF9 Control? Only the depositors can answer, and the answer decides which arm three of this beat's tracks belong to.
- How many other filename collisions exist across the 158 samples? Nothing has swept them.
the control that licenses Task 1's pairing as within-arm rather than cross-arm
a second way this deposit's file metadata misleads about which samples are comparable
supplies the calibration range without which C-IDENT's 0.83 could not be read