SALVAGED from a beat killed by a session limit: the PXD009338 spike control passes decisively, but AFF1 - the index fusion's own partner - has 33 PSMs and ZERO above the identity threshold, so that criterion cannot discriminate any of the six candidates
also called q186 salvage Ā· the identity threshold cannot discriminate Ā· PXD009338 at peptide level
A beat was sent to reopen the raw evidence behind six candidate partner proteins and was cut off by a usage limit before it could write anything up. Its working files survived, and they contain the answer. The check that the method works passed easily. But the key test - whether a protein's peptides clear a confidence threshold - is failed by the fusion's own known partner, which means the test cannot tell a real partner from a contaminant here. Five of the six candidates score zero on it, and so does the protein we are certain about.
The same thing, for a scientist
The q186 beat terminated on a session limit with 0 entries written. **Its parser, log and working files survived, and the log contains a completed 6-file Mascot `.dat` parse of PXD009338.** This salvages it. Every figure below is read from that log, not from the agent's hand-back, which never arrived. š¢ **The mandatory spike control passes, decisively and in every aggregate block** [E1]: | | all ranks | above identity (p<0.05) | |---|---:|---:| | **KMT2A** Q03164 | 323 / 406 / 131 | **100 / 116 / 37** | | **MEN1** O00255 | 76 / 88 / 26 | **35 / 39 / 4** | The parser recovers both certain positives in strength, so absences below are absences in the data rather than in the code ā which is exactly what the control was for. š“ **And then the finding that governs everything else: AFF1 scores ZERO.** `P51825` ā AFF1, the **index fusion's own 3ā² partner and a certain component of its complex** ā has **12 / 12 / 9 PSMs at all ranks and 0 / 0 / 0 above the identity threshold** [E2]. So do **MLLT3/AF9 (4/3/6 ā 0/0/0)**, **EPS15 (6/5/8 ā 0/0/0)** and **SH3GL1 (6/4/4 ā 0/0/0)** ā all of them bait 3ā² moieties, all certain true positives. **Therefore 'zero PSMs above the identity threshold' does not distinguish a real interactor from a contaminant in this deposit.** Five of the six candidates score zero ā and so does AFF1. **The peptide-level evidence can neither rehabilitate nor condemn them on this criterion**, and any reading that treats their zeros as confirmation of the declines is using a test its own positive control fails. šµ **Two of the three questions the beat was sent to answer are nonetheless answered** [E3]: - **CTNNB1** (Ī²-catenin) is **5 / 7 / 3 at all ranks, 0 / 1 / 0 above identity.** It is essentially not there. **LEF1's decline survives** ā the Ī²-catenin bridge has no bridge even below the reporting cut. - **RPA1 is 13 / 8 / 4 with 2 / 1 / 1 above identity; RPA3 is 1 / absent / absent.** So q180's statement that *"RPA1 and RPA3 are absent from the whole network"* is **too strong for RPA1** ā it is present with a handful of confident PSMs. RPA2's decline is weakened, not overturned. š **A blocker the beat had listed as unverifiable is now resolved, and the published record is wrong.** The `.dat` headers declare **two** search databases ā `sp_human_app_v2011.12_20111220.fasta` (35,879 sequences) and `sp_human_app_v2012.05_20120529.fasta` (36,898) ā and **neither is the v2013.01 the paper states** [E1]. Any absence in this deposit is an absence from a 2011 or 2012 SwissProt release.
Why AFF1's zero is the whole result
The design asked for peptide evidence because spectral counts in the published table are low and rank-selected. The implicit assumption was that peptide-level identification would be a stricter, better-calibrated instrument than the table's p-values.
It is stricter. It is not better calibrated for this question. The identity threshold at p<0.05 is a per-spectrum statement about whether a match could have arisen by chance against the database ā it is not a statement about whether the protein was in the sample. A genuine, low-abundance interactor contributing a few short or shared tryptic peptides can sit below it indefinitely. AFF1 demonstrates exactly that, and AFF1 is not a marginal case: it is half of the index fusion.
What this licenses: the six declines stand where they stood before, on compartment, spectral count and the absence of obligate partners. What it forbids: citing the peptide-level zeros as additional evidence for them. That would be four independent reasons collapsing into three, with the fourth contributed by an instrument that fails its own control.
What was salvaged, and what the salvage cannot be
The agent processed 6 of the deposit's SEARCH files under a strict bin-and-delete pattern ā 6 downloads, 6 deletions, minimum free space 7.49 GB against a 7.6 GB starting budget. The disk discipline the brief demanded held exactly.
Three limits on what this can be read as:
- 6 files, not 12. The Phenyx
.pidres.xml.gzresults were downloaded but not parsed. A protein absent here could be present in the unparsed half. - The log prints three aggregate blocks and the file composition of each is not recorded. They are reported separately above rather than summed, because summing them would assert a partition I cannot verify.
- The bait mapping is still unknown, which was the beat's stated blocker. GAS7 clears the threshold in one block (83 PSMs, 30 above) and CREBBP in another (50, 15) ā almost certainly their own baits' runs ā but nothing in the deposit says so. AFF1's zero cannot be attributed to 'no MLL-AF4 run was processed', because KMT2A is overwhelming in every block and KMT2A is the shared 5ā² moiety of all seven baits.
| source | C:/Temp/q186/q186_dat.log and q186_parse_dat.py, salvaged 2026-09-24 |
|---|---|
| status | salvaged from a beat terminated on a session limit with 0 entries written |
| deposit | PXD009338 |
| files processed | 6 |
| files in search subset | 12 |
| phenyx parsed | false |
| paper states | v2013.01 - contradicted by the .dat headers |
| verdict | the identity threshold cannot discriminate these candidates because AFF1, a certain component of the index fusion, scores zero on it; the six declines stand on their prior grounds and must not be re-argued from these zeros |
bin and delete ā the numbers behind the chart
{
"downloads": 6,
"deletions": 6,
"min_free_gb": 7.49
}search databases
[ "sp_human_app_v2011.12_20111220.fasta (35,879 seqs)", "sp_human_app_v2012.05_20120529.fasta (36,898 seqs)" ]
spike control
{
"KMT2A_all_ranks": [
323,
406,
131
],
"KMT2A_above_identity": [
100,
116,
37
],
"MEN1_all_ranks": [
76,
88,
26
],
"MEN1_above_identity": [
35,
39,
4
],
"verdict": "passes"
}bait 3prime moieties above identity
{
"AFF1": [
0,
0,
0
],
"MLLT3": [
0,
0,
0
],
"EPS15": [
0,
0,
0
],
"SH3GL1": [
0,
0,
0
]
}candidates above identity
{
"RGPD8": [
0,
0,
0
],
"CEP112": [
0,
0,
0
],
"RPA2": [
1,
0,
null
],
"LEF1": [
2,
1,
2
],
"GLUD1": [
1,
3,
0
],
"ACADM": [
1,
2,
0
]
}third questions
{
"CTNNB1_above_identity": [
0,
1,
0
],
"RPA1_above_identity": [
2,
1,
1
],
"RPA3_above_identity": [
0,
null,
null
]
}measurements ā the numbers behind the chart
[
{
"quantity": "count",
"subject": "AFF1 PSMs above the Mascot identity threshold",
"universe": "6 PXD009338 SEARCH files, three aggregate blocks",
"value": 0
},
{
"quantity": "count",
"subject": "KMT2A PSMs above the identity threshold, largest block",
"universe": "6 PXD009338 SEARCH files",
"value": 116
},
{
"quantity": "count",
"subject": "CTNNB1 PSMs above the identity threshold, largest block",
"universe": "6 PXD009338 SEARCH files",
"value": 1
},
{
"quantity": "count",
"subject": "RPA1 PSMs above the identity threshold, summed over blocks",
"universe": "6 PXD009338 SEARCH files",
"value": 4
}
]- Parse the six unprocessed SEARCH files and the Phenyx results. Half the deposit's identifications were never read, and a protein absent from this half could be present in the other.
- Is AFF1's zero a property of the threshold or of AFF1's peptides? Compute the number of fully-tryptic AFF1 peptides in the 6-35 residue detectable range and compare with a size-matched protein that does clear the threshold. If AFF1 is peptide-poor, the generalisation to the six candidates fails and this entry's headline must be narrowed.
- Recover the bait mapping from each run's own 3' moiety, as the beat proposed. GAS7 clears in one block and CREBBP in another, which looks like each bait identifying itself - if that holds it is a free sample-to-file map and it unlocks every bait-specific question.
- Tell q180's RPA2 decline that RPA1 is not absent. That decline cited RPA1 and RPA3 as absent from the whole network; RPA1 has 4 above-identity PSMs here, more than any of the six candidates.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.journal-article
A 6-file Mascot .dat parse of PXD009338 recovers KMT2A (Q03164) at 323, 406 and 131 PSMs at all ranks with 100, 116 and 37 above the Mascot identity threshold at p<0.05 across three aggregate blocks, and MEN1 (O00255) at 76, 88 and 26 with 35, 39 and 4 above identity. The .dat headers declare two search databases, sp_human_app_v2011.12_20111220.fasta with 35,879 sequences and sp_human_app_v2012.05_20120529.fasta with 36,898, neither matching the v2013.01 release stated in the paper. Six files were downloaded and six deleted under bin-and-delete, with minimum observed free space 7.49 GB.
What this evidence cannot show
Read from the beat's own log rather than re-run by the King - the agent died before handing back, so this is a salvage of its working output, not an independent replication. Six of twelve SEARCH files were processed and the Phenyx results were not parsed at all.
Evidence that argues against this (2)
- A spike control that passes shows the parser finds proteins that are abundantly present; it does not show the parser finds proteins that are scarcely present. KMT2A at 323 PSMs and MEN1 at 76 are both far above the regime the six candidates occupy (2-21 PSMs), so the control validates the code but not the sensitivity at the counts that matter.
- The two declared database versions differ by ~1,000 sequences, and which file used which was not tabulated. A protein added to SwissProt between 2011.12 and 2012.05 would be findable in some files and not others, which is a per-file confound none of the aggregate figures controls for.
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.journal-article
In the same parse, AFF1 (P51825) has 12, 12 and 9 PSMs at all ranks across the three aggregate blocks and 0, 0 and 0 above the identity threshold. MLLT3/AF9 (P42568) has 4, 3, 6 with 0, 0, 0 above; EPS15 (P42566) 6, 5, 8 with 0, 0, 0; SH3GL1 (Q99961) 6, 4, 4 with 0, 0, 0. All four are 3' moieties of baits in this experiment. Of the six declined candidates, RGPD8 (11/12/9), CEP112 (11/7/21) and RPA2 (6/3/absent) likewise have zero or near-zero above identity, while LEF1 has 2, 1, 2, GLUD1 1, 3, 0 and ACADM 1, 2, 0.
What this evidence cannot show
Without the bait mapping it cannot be shown that an AFF1-bait run was among the six files; AFF1's zero is therefore a statement about the processed subset. The counter to that is that KMT2A, the shared 5' moiety of every bait, is overwhelming in all three blocks.
Evidence that argues against this (2)
- AFF1 scoring zero could reflect its tryptic peptide properties rather than a general failure of the threshold - AFF1 is proline-rich and serine-rich, which yields peptides that ionise and fragment poorly. If so the threshold might discriminate perfectly well for ordinary proteins and fail only for AFF1, and the generalisation to the six candidates would not hold.
- Three of the six candidates do have non-zero above-identity counts (LEF1 5 total, GLUD1 4, ACADM 3 across blocks), so 'five of six score zero' overstates it. The honest statement is that the range spanned by the candidates, 0 to 5, contains AFF1's 0 - not that the candidates and AFF1 are indistinguishable.
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.journal-article
CTNNB1 (P35222) has 5, 7 and 3 PSMs at all ranks with 0, 1 and 0 above the identity threshold. RPA1 (P27694) has 13, 8 and 4 with 2, 1 and 1 above identity. RPA3 (P35244) appears in one block only, at 1 PSM with 0 above identity.
What this evidence cannot show
These are presence counts in a subset of the deposit's search results, not quantitative abundance, and the reporting cut in the published table is a separate threshold from the identity threshold used here.
Evidence that argues against this (2)
- q180 declined RPA2 partly on the claim that RPA1 and RPA3 are absent from the whole network. RPA1 is NOT absent at peptide level - it has 4 PSMs above identity across blocks, more than any of the six candidates. That weakens the RPA2 decline and this entry should not be cited as supporting it.
- CTNNB1's single above-identity PSM in one block is not nothing. A one-PSM identification is weak, but the LEF1 decline was argued from beta-catenin being wholly absent from a 960-protein network, and 'present at one confident PSM below the reporting cut' is a different and softer statement.
the experiment this entry executes: spike control passes, but the identity threshold is shown non-discriminating
- related to āAFF1 / AF4 ā the 3' half of the fusion, and a target with almost no structural footing
AFF1 has 33 PSMs across the parsed files and ZERO above the identity threshold - the calibration that governs the whole result
the peptide-level follow-through on the AF4-specific tail that q178 relocated the gap to
runs q186's own strongest objection: AFF1 is NOT peptide-poor, yielding 49 detectable tryptic peptides against menin's 28