PSIP1-FOXK2: a single TAP row from the same TF atlas as the RBPJ decline - kept as a lead rather than declined, because FOXK2 is a chromatin-recruiting forkhead expressed in lymphoid cells and LEDGF is a chromatin-recruiting tether
also called LEDGF-FOXK2 · PSIP1-FOXK2 · forkhead box K2
FOXK2 is a DNA-binding protein that brings other complexes onto chromatin, and LEDGF does the same job for the MLL machinery. They came up together in one big survey, which is thin evidence for two proteins that do similar things.
The same thing, for a scientist
**Verdict: LEAD, not declined - and the difference from the sibling RBPJ entry is deliberate and narrow.** IntAct holds one PSIP1-FOXK2 record: **TAP**, *association*, HEK293T, PSIP1 as prey, intact-miscore 0.35, from **PMID 25609649** [E1] - the same 214-purification, 2,156-interaction transcription-factor TAP atlas this beat's atlas heuristic failed to flag. **On the record, FOXK2 and RBPJ are identical rows from the same study.** They are graded differently on one thing only: **mechanism compatibility**. FOXK2 (Q01167, 660 aa) is a **forkhead transcription factor with four experimentally coded nuclear localization entries and tissue specificity recorded as expressed in both lymphoid and non-lymphoid cells** [E2], and its documented function is **recruiting chromatin-modifying complexes to DNA**. PSIP1/LEDGF's documented function in this disease is **tethering the KMT2A-menin module to chromatin through its PWWP domain**. **Two tethers recovered together is a more specific hypothesis than a tether recovered with a pathway effector, and the atlas's own design - separating chromatin-associated from soluble complexes - is the one that could test it.**
Why this is a lead and RBPJ is a decline, when the rows are identical
The rows are the same. The grading difference is a prior, and it is stated as one rather than dressed as evidence.
| RBPJ | FOXK2 | |
|---|---|---|
| record | 1 TAP, HEK293T, 0.35, PMID 25609649 | 1 TAP, HEK293T, 0.35, PMID 25609649 |
| nuclear, experimentally coded | yes | yes, four codes |
| lymphoid expression | yes | yes, explicitly 'lymphoid and non-lymphoid' |
| mechanism overlap with PSIP1's known role | no - Notch pathway effector | yes - both are chromatin-recruitment factors |
| why it matters | Notch is the dominant pathway of an adjacent leukaemia, which makes a weak row dangerous to promote | chromatin recruitment is the exact function LEDGF performs for the fusion |
The asymmetry is honest and it is also the thing to be suspicious of. A prior that promotes rows compatible with the mechanism we already believe is how a corpus confirms itself. The protection is that both entries exist, the rows are declared identical, and the grading criterion is written down where it can be disagreed with - rather than one row being written up and the other silently dropped.
Neither is evidence of anything yet. A lead and a decline on identical rows differ in what work they invite, not in what they establish.
The experiment the source already ran and IntAct did not record
PMID 25609649's headline is that transcription factors form distinct protein complexes on and off chromatin, and it purified soluble and chromatin-associated fractions separately across 214 purifications 3.
So the study already knows whether the PSIP1 recovery came from the chromatin fraction or the soluble one, and IntAct's curation does not carry it. For a pair of putative chromatin-recruitment factors that is the single most informative fact available, and it costs one supplementary table to retrieve.
This is a concrete, specific gap in interaction-database curation rather than a complaint about atlases. The PSI-MI vocabulary has no widely used field for the biochemical fraction a co-purification came from, so a paper whose entire point is fraction-dependence is curated as though fraction did not exist. Every row from this study in this census - RBPJ, FOXB1, FOXE1, FOXK2, JUN - loses the same information.
What would make FOXK2 an admission: the recovery being in the chromatin fraction, plus a reciprocal or an endogenous co-IP. The first is retrievable today from the paper; the second does not exist.
| partner a | gene-psip1 |
|---|---|
| partner b | FOXK2 |
| partner b uniprot | Q01167 |
| partner b length AA | 660 |
| partner a uniprot | O75475 |
| partner a length AA | 530 |
| verdict | LEAD |
| adjudicated by | q160, cycle 146, 2026-09-24 |
| druggability | Not assessable. Neither protein has a chemical probe in this corpus; LEDGF's PWWP-DNA interface is the nearest tractable object and is not implicated here. |
| n methods | 1 |
| n platforms | 1 |
| n publications | 1 |
| endogenous side | no - tagged TF bait |
| intact miscore | 0.35 |
| cell state | the source separates chromatin-associated from soluble complexes; IntAct does not record which fraction this row came from |
| epigenetics | FOXK2's documented role is recruiting chromatin-modifying complexes to DNA; PSIP1/LEDGF's role in this disease is tethering the KMT2A-menin module to chromatin via its PWWP domain. No measurement connects the two. |
| grading criterion | Graded LEAD rather than DECLINED on a stated PRIOR about mechanism compatibility, not on the record. The sibling PSIP1-RBPJ row from the same publication is identical and is declined. |
| absence class | measured-not-analysed |
| absence class note | UNUSUAL FOR THIS CORPUS: the discriminating measurement EXISTS. PMID 25609649 separated chromatin-associated from soluble complexes and its supplementary tables record which fraction each recovery came from. Retrieving that for the PSIP1 rows requires no new experiment and was not done here. |
interface residues a
[]
structures
[]
ligands
[]
records
[ "PSIP1-FOXK2, tap, association, HEK293T, PSIP1 prey, miscore 0.35, PMID 25609649" ]
cell systems
[ "HEK293T embryonic kidney" ]
localization
{
"compartment": "nucleus",
"subnuclear": [],
"evidence_codes": "four x ECO:0000269",
"shown_for_fusion": "NO - nobody has looked in any fusion."
}measurements
[
{
"quantity": "count",
"subject": "prey-only rows in this census from PMID 25609649",
"value": 5,
"units": "rows",
"n": 212,
"universe": "the 2026-09-24 IntAct REST prey-only census across the six baits",
"null": 15,
"note": "RBPJ, FOXB1, FOXE1, FOXK2, JUN. All five lose the chromatin/soluble fraction label in curation."
}
]- Promoting a row because it is compatible with the mechanism we already believe is the mechanism by which a knowledge graph confirms itself. This entry states the prior explicitly and pairs itself with an identical row that is declined, which is the only defence available short of not writing it.
- Did the PSIP1 recovery in PMID 25609649 come from the chromatin-associated or the soluble fraction? This is retrievable from that paper's supplementary tables today, it decides whether this entry is a lead or a decline, and it applies identically to the RBPJ, FOXB1, FOXE1 and JUN rows.
- Should IntAct and the PSI-MI vocabulary carry a biochemical-fraction field? A study whose headline is fraction-dependence is curated as though fraction did not exist, and every row from it loses the same information.
- Does LEDGF co-occupy chromatin with FOXK2 genome-wide? Public LEDGF and FOXK2 ChIP datasets exist separately and have never been intersected.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
IntAct REST returns one PSIP1-FOXK2 (O75475-Q01167) record: detection method 'tap', interaction type 'association', host Homo sapiens HEK293T embryonic kidney cell, PSIP1 as PREY, intact-miscore 0.35, from PMID 25609649. The PSIP1-RBPJ record from the same publication is identical in method, type, host, role and score.
IntAct molecular interaction database, REST findInteractions on O75475, retrieved 2026-09-24 by q160 (partner FOXK2) · IntAct (EMBL-EBI) 2026What this evidence cannot show
One record. PSIP1 is an abundant chromatin-tethered protein and a frequent recovery in nuclear purifications; this corpus records it as prey to nine of fifty RBP baits in an unrelated map, which is the base rate this row sits against and which no calculation here adjusts for.
Evidence that argues against this (1)
- The RBPJ row is identical and is declined. If the two rows are the same and the verdicts differ, then the verdict is carried entirely by a prior about mechanism, and a reader who does not share that prior should read both as declines. IntAct (EMBL-EBI) 2026
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
UniProtKB Q01167 (FOXK2_HUMAN) is 660 aa, reviewed. Its subcellular location block records Nucleus with FOUR experimental evidence codes (ECO:0000269) and Cytoplasm by similarity (ECO:0000250). Tissue specificity: expressed in both lymphoid and non-lymphoid cells.
UniProtKB entry Q01167, retrieved 2026-09-24 · UniProt Consortium 2026What this evidence cannot show
Rule 91: four ECO:0000269 codes mean four curated experiments, and none of the cited papers was opened here, so it is not known whether any of them is a haematopoietic cell. 'Expressed in both lymphoid and non-lymphoid cells' removes an expression-mismatch objection; it supplies no positive evidence in a leukaemic blast.
Evidence that argues against this (1)
- Four nuclear localization codes and broad expression are exactly what an abundant, ubiquitous nuclear transcription factor looks like, and abundance is the leading cause of spurious AP-MS recovery. Every property cited here as compatibility is also a property that raises the background rate. UniProt Consortium 2026
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
PMID 25609649 (Li et al., Mol Syst Biol 2015) performed 214 tandem-affinity purifications across 56 transcription factors including 37 Forkhead-box family members, separating SOLUBLE and CHROMATIN-ASSOCIATED complexes, and identified 2,156 high-confidence protein-protein interactions, concluding that most transcription factors form very distinct protein complexes on and off chromatin. IntAct's curation of the PSIP1-FOXK2 row does not record which fraction it came from.
Proteomic analyses reveal distinct chromatin-associated and soluble transcription factor complexes. · Li X, Wang W, Wang J, Malovannaya A, Xi Y, Li W, Guerra R, Hawke DH, Qin J, Chen J · Mol Syst Biol 2015What this evidence cannot show
Abstract read in full; the supplementary interaction tables, which would say which fraction the PSIP1 recovery came from, were NOT retrieved. That is the single most informative missing fact in this entry and it is retrievable without new experiments.
Evidence that argues against this (1)
- If the PSIP1 recovery is from the SOLUBLE fraction, the chromatin-recruitment rationale that makes this a lead rather than a decline collapses entirely, and this entry should be rewritten as a decline. The entry is written knowing that the fact which could kill it was available and not fetched. Li X et al. 2015
single TAP row, HEK293T; graded LEAD on a stated mechanism prior, not on the record
identical row from the same publication, graded DECLINE - the pair is written so the prior can be disagreed with
one of the 135 uncovered human prey-only partners