AFF1 is NOT peptide-poor - it yields 49 detectable tryptic peptides against menin's 28, and menin cleared the identity threshold while AFF1 scored zero - but the two groups do not separate on peptide yield at all, so the driver is abundance
also called q189 · is AFF1 peptide-poor · the third option again
Last cycle found that the fusion's own partner protein scores zero on a confidence test that other proteins pass, and concluded the test cannot be trusted here. The obvious objection, which that entry raised against itself, is that this particular protein might simply be hard to detect - it is unusually rich in two amino acids that can make protein fragments awkward to measure. Cutting all ten proteins up in software and counting the measurable fragments: it is not hard to detect. It yields more measurable fragments than menin, which passes the test easily. But the proteins that pass and the proteins that fail are indistinguishable on this measure, so fragment yield explains nothing either way.
The same thing, for a scientist
Cycle 160's salvage concluded that the Mascot identity threshold cannot discriminate in PXD009338, because **AFF1 — the index fusion's own 3′ partner — has 33 PSMs and zero above it** while KMT2A and MEN1 clear it overwhelmingly. **That entry recorded its own strongest objection**: AFF1 is proline- and serine-rich, so its tryptic peptides might simply be undetectable, in which case the generalisation to six declined candidates fails. This tests it. 🟢 **The compositional worry is real and its predicted consequence does not follow.** AFF1 is **11.2% proline** (highest but one of the ten, behind CREBBP's 11.3%) and **16.3% serine** (second only to MLLT3's 20.1%) [E1]. But a fully-tryptic digest counting peptides in the 7–30 residue detectable window gives AFF1 **40.50 per 1000 residues — squarely mid-range**, with **3 of the 6 threshold-clearing proteins above it and 3 below** [E1]. 🔴 **The decisive pair: AFF1 yields 49 detectable peptides; MEN1 yields 28.** MEN1 cleared the threshold with **35, 39 and 4 PSMs above identity** across the three blocks. **AFF1, with 75% more detectable peptides available, scored zero.** Peptide availability is not the explanation [E2]. 🟠 **But the third pre-registered option is what actually realised, and it constrains the conclusion.** The cleared and zero groups **do not separate on peptide density at all**: cleared median **42.91**, zero median **45.92**, ratio **0.93** — the group that failed is marginally *denser* [E2]. **So tryptic yield explains nothing in either direction.** **What this licenses:** cycle 160's headline survives the specific objection it raised against itself. AFF1's zero is not peptide poverty. **What it forbids:** concluding that the threshold is therefore broken. **The remaining explanation is abundance — how much of each protein was in the pulldown — and this design cannot measure it.** The honest position is that the identity threshold does not discriminate *here*, for a reason still unidentified. 🔵 **Pre-registering the third option is what produced this.** Written before any sequence was fetched: *"the cleared and zero groups may not separate on density at all, in which case peptide yield explains nothing either way and the real driver is abundance — which this design cannot measure and should not pretend to."* Without it, the AFF1-vs-MEN1 pair would have been read as proof that the threshold is broken.
Why the AFF1-versus-MEN1 pair is the part that carries weight
Group medians on ten proteins are weak — four in one arm, six in the other, and the arms overlap completely. The single paired comparison is much stronger, because it removes the grouping entirely:
| detectable peptides | PSMs above identity | |
|---|---|---|
| MEN1 | 28 | 35 / 39 / 4 |
| AFF1 | 49 | 0 / 0 / 0 |
Both are certain components of the complexes being pulled down. One has fewer peptides available and is identified confidently in every block; the other has 75% more and is identified confidently in none. No property of the digest can produce that ordering.
The caveat that matters: MEN1 is 610 residues and AFF1 is 1,210, so AFF1's higher absolute count partly reflects being twice the size. Per 1000 residues they are nearly identical — 45.90 for MEN1, 40.50 for AFF1 — which is the point: on a per-residue basis they are the same kind of substrate, and they behave completely differently in the data.
What the design cannot see, stated plainly
Abundance. An in-silico digest counts what peptides could be produced; it says nothing about how many molecules of the protein were in the tube. If AFF1 is present at a tenth of MEN1's molar abundance in these pulldowns, everything observed follows without any instrument defect. This analysis cannot distinguish that from a threshold problem, and the entry it supports should not be read as having done so.
Ionisation and fragmentation. Peptide length is a crude proxy for detectability. Hydrophobicity, charge state, missed cleavages and post-translational modification all shift which peptides are actually observed, and none is modelled here. A retention-time or ionisation-efficiency prediction would be a better instrument and was not run.
Ten proteins. Six cleared, four zero. That is far too few to establish that the groups never separate on density — only that they do not separate in this comparison, at medians 42.91 against 45.92.
| script | analyses/q189-is-aff1-peptide-poor-or-is-the-threshold-broken/run.py |
|---|---|
| status | executed |
| answers open question from | ana-q186-salvaged-the-identity-threshold-cannot-discriminate-because-aff1-itself-scores-zero |
| digest | fully tryptic, cleave after K/R not before P, no missed cleavages, 7-30 aa window |
| cleared median det per 1000 | 42.91 |
| zero median det per 1000 | 45.92 |
| ratio cleared over zero | 0.934 |
| cleared with higher density than aff1 | 3 |
| cleared total | 6 |
| preregistered outcome realised | 3 |
| verdict | AFF1 is not peptide-poor and the cycle-160 objection fails; but the groups do not separate on yield, so the threshold is not thereby shown broken and the remaining explanation is abundance, which this design cannot measure |
aff1 — the numbers behind the chart
{
"length": 1210,
"n_detectable": 49,
"det_per_1000": 40.5,
"pct_P": 11.2,
"pct_S": 16.3,
"pct_KR": 14,
"pct_seq_covered": 51.3
}men1 — the numbers behind the chart
{
"length": 610,
"n_detectable": 28,
"det_per_1000": 45.9
}measurements — the numbers behind the chart
[
{
"quantity": "count",
"subject": "AFF1 detectable fully-tryptic peptides, 7-30 aa",
"universe": "UniProt P51825, 1210 aa",
"value": 49
},
{
"quantity": "count",
"subject": "MEN1 detectable fully-tryptic peptides, 7-30 aa",
"universe": "UniProt O00255, 610 aa",
"value": 28
},
{
"quantity": "ratio",
"subject": "median detectable-peptide density, threshold-clearing over threshold-failing proteins",
"universe": "6 cleared and 4 zero proteins from PXD009338",
"value": 0.934
},
{
"quantity": "density",
"subject": "AFF1 detectable peptides per 1000 residues",
"universe": "10-protein comparison set",
"value": 40.5
}
]- Measure abundance, which is the only surviving explanation. PaxDb or a whole-cell proteome of a KMT2A-rearranged line would give molar estimates for AFF1 and MEN1; if AFF1 is an order of magnitude scarcer, no instrument defect is needed and cycle 160's framing should be replaced.
- Re-run the digest allowing 1-2 missed cleavages and with an ionisation-efficiency or retention-time prediction. Peptide length is a crude proxy and AFF1's high K+R content means the no-missed-cleavage assumption may understate its yield.
- Settle whether AFF1's own bait run was among the six SEARCH files parsed. If it was not, AFF1's zero is a sampling artefact and both this entry and the cycle-160 salvage are answering the wrong question. Recovering the bait mapping from each run's 3' moiety would decide it.
- Extend the comparison beyond ten proteins. Six versus four with overlapping ranges cannot establish that peptide density never separates identified from unidentified proteins; the full 960-protein network with per-protein PSM counts would make it a real test.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
In-silico fully-tryptic digestion (cleave after K or R, not before P, no missed cleavages) of ten UniProt sequences fetched from the REST API, counting peptides 7-30 residues long: AFF1 (P51825, 1210 aa) yields 155 peptides of which 49 are in the detectable window, 40.50 per 1000 residues, covering 51.3% of the sequence. Threshold-clearing proteins give DDX5 53.75, DDX17 48.01, MEN1 45.90, GAS7 39.92, KMT2A 38.55 and CREBBP 30.30 per 1000 residues; three of the six are above AFF1 and three below. AFF1 is 11.2% proline and 16.3% serine, the second-highest of each in the set.
What this evidence cannot show
A fully-tryptic digest with no missed cleavages is an idealisation; real searches allow 1-2 missed cleavages, which would change every count though not obviously the ordering. The 7-30 residue window is a convention, not a property of this instrument, and was not swept.
Evidence that argues against this (2)
- Peptide length is a weak proxy for detectability. Hydrophobicity, charge state and ionisation efficiency determine what is actually observed, and none is modelled. A serine-rich protein could yield many length-appropriate peptides that ionise poorly, which would rescue the peptide-poverty explanation this entry rejects.
- Allowing missed cleavages would help AFF1 disproportionately if its K/R residues are clustered, since clustered sites produce many too-short fully-tryptic peptides that become detectable only when a cleavage is missed. AFF1's K+R content is 14.0%, among the highest in the set, so the no-missed-cleavage assumption may understate its real yield - which would strengthen this entry's conclusion, not weaken it.
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
The six proteins that cleared the identity threshold in PXD009338 have a median detectable-peptide density of 42.91 per 1000 residues; the four that scored zero have 45.92; the ratio is 0.93. AFF1 yields 49 detectable peptides against MEN1's 28, while MEN1 was identified at 35, 39 and 4 PSMs above the identity threshold across three aggregate blocks and AFF1 at 0, 0 and 0.
What this evidence cannot show
Six versus four proteins, with completely overlapping ranges - the medians are descriptive, not a test. The PSM counts are taken from the cycle-160 salvage of another agent's log, not re-derived here.
Evidence that argues against this (2)
- A ratio of 0.93 on groups of six and four is indistinguishable from 1.0 and also indistinguishable from a real difference of 20% in either direction. Reporting 'the groups do not separate' overstates what ten points can show; the honest statement is that no separation is detectable at this n.
- The whole comparison assumes the cleared/zero split is a property of the proteins rather than of which files were processed. Only 6 of 12 SEARCH files were parsed, and GAS7 clears in one block while CREBBP clears in another - consistent with each bait identifying itself. If AFF1's own bait run was not among the six files, its zero is a sampling artefact and this entire analysis is answering the wrong question.
runs q186's own strongest objection: AFF1 is NOT peptide-poor, yielding 49 detectable tryptic peptides against menin's 28
AFF1's tryptic yield is mid-range at 40.50 detectable peptides per 1000 residues despite being 11.2% proline and 16.3% serine
menin yields 28 detectable peptides and clears the identity threshold; AFF1 yields 49 and scores zero