PA2G4 / EBP1 (Q9UQ80) - the best-evidenced member of the MLL-fusion nucleolar block in the index fusion's own column (p = 0.014), and the only one with a published AML phenotype

PA2G4 IS in the MLL-AF4 column, at combined p = 0.014, recovered by 6 of the seven baits (MLL-AF1p, MLL-AF4, MLL-AF9, MLL-EEN, MLL-ENL, MLL-GAS7). Negative control: ZERO spectra. This is the LOWEST MLL-AF4 p-value of the eleven nucleolar-block proteins this beat wrote, and one of only four block members below 0.05 in that column. FUNCTIONAL SCREEN. Supplementary Data 2 (MOESM5) is the shRNA library for the negative-selection screen: 768 data rows, 128 unique gene symbols, EXACTLY six shRNAs each -- counted with openpyxl by this beat on 2026-09-21, and this protein's symbol IS in it with six shRNAs. So it was functionally tested against an MLL fusion (MLL-AF9 in MOLM-13), which is more than a co-purification. But per-gene scores are published only as the heatmaps in Fig. 2c/2d with no table anywhere in the paper or its five supplementary files, so the outcome for this protein is not recoverable: measured-not-reported. THE BLOCK, re-measured by this beat on 2026-09-21 by re-running analyses/mll-fusion-interactome-complexes/run.py. A 26-subunit nucleolar / ribosome-biogenesis set returns 15 of 26 present in the 960-protein network against 1.25 expected by chance, hypergeometric p = 6.99e-14, q(BH) = 8.39e-13 across the twelve complexes tested -- the TOP-RANKED complex in the whole interactome, ahead of the CCT/TRiC negative control (q = 1.4e-09) and the SEC positive control (q = 1.5e-09). The fifteen are FBL, FTSJ3, GNL2, LYAR, NOLC1, NOP2, NOP58, PA2G4, RRP12, RRP9, TBL3, TCOF1, UTP14A, UTP18, WDR18. BUT THE BACKGROUND IS THE ~20,000-PROTEIN PROTEOME, not the ~4,600 proteins the study actually recovered before filtering, so every q here is an UPPER BOUND on significance and it is loosest for a complex of abundant nuclear proteins -- which this one is, more than any other row in that table. AND THE POOLED NUMBER IS NOT THE PER-BAIT NUMBER. Subunits of the block recovered per bait, out of 26: MLL-AF1p 8, MLL-AF4 7, MLL-AF9 12, MLL-CBP 10, MLL-EEN 12, MLL-ENL 11, MLL-GAS7 7. The index fusion is TIED LAST. In its column there are seven members and only four at p<0.05 -- GNL2, NOP2, PA2G4 and TBL3; LYAR (0.141), NOLC1 (0.134) and TCOF1 (0.163) are above it. So the strongest complex-level signal in this entire interactome is weakest in the fusion this project studies. SYSTEM: the AP-MS host is JURKAT, a human T-ALL line, while KMT2A::AFF1 disease is B-lineage infant leukaemia; a partner requiring a B-cell or infant proteome cannot appear here and a T-lineage partner can appear spuriously. The bait is a Strep-HA-tagged, SINGLE-COPY, doxycycline-induced transgene at an FRT/Flp cassette, not the endogenous fusion. Prey came from bulk nuclear lysate after a single STREP-Tactin step, so co-purification is NOT evidence of a direct contact. INCLUSION RULE: the paper kept the top 300 interactions per bait by combined p-value rather than applying a p threshold (Methods, verbatim: "All the identified bait-prey pairs were ranked by the combined p-value. The 300 most significant interactions per bait were retained."). In the MLL-AF4 column 186 of 291 listed p-values (63.9%) exceed 0.05; the pooled rate over all baits is 26.6%, and the pooled rate must NOT be quoted for a per-bait claim. BAIT DEPTH: MLL-AF4 is the shallowest of the seven baits by spectral counts, so a low or absent MLL-AF4 count is a property of the bait (transgene expression, purification depth, normalisation) and must not be read as a weaker association. ABUNDANCE IS THE LEADING ALTERNATIVE EXPLANATION. The nucleolus is the densest compartment in the nucleus, every bait here is a tagged nuclear protein, and the complex-level analysis in this repo exists to make exactly that point: its strongest per-bait signature in the index fusion's own column is CCT/TRiC, a canonical AP-MS contaminant, at 8 of 9 subunits with 6 at p<0.05 -- BETTER represented in MLL-AF4's column than the nucleolar block is. Complex-wide enrichment is not by itself evidence of specific binding. THE ARGUMENT ON THE OTHER SIDE: all thirteen members of the nucleolar block, this protein included, have a negative control of ZERO spectra. A non-specific sticky abundant protein would be expected to appear in the control purification too, and none of them does. That is the strongest single fact about this block and it is not explained by abundance alone. SELECTIVITY: the paper's own named NON-selectivity controls are RFC1, SF3A3 and CCT3 -- knockdown inhibited MLL-wild-type K562 and HL-60 as much as MOLM-13 -- against MEN1 as the selective positive. This protein is not among them. That is NOT evidence of selectivity: only 40 of the 128 entered the MLL-r versus MLL-wild-type re-screen and the paper does not say which 40, so absence from the control list is equally consistent with never having reached the second screen. Recorded as UNKNOWN. This entry does not nominate a target. Per-bait membership re-derived by this beat on 2026-09-21 by re-running analyses/skucha-mll-fusion-interactome/run.py (both controls PASS: 128 shared by >=5 baits, 42 by all seven, MEN1 present with all seven) and analyses/mll-fusion-interactome-complexes/run.py. n_fusions and the MLL-AF4 p-value were supplied pre-verified by the coordinator from the cached MOESM4 xlsx and were independently reproduced by this beat's re-run of the complex analysis.

• IntAct has never curated this dataset, so no independent curator has re-examined these calls and no interactome-coverage metric built on IntAct can see them; the raw spectra at ProteomeXchange PXD009338 have not been reanalysed here. The authors themselves removed seven proteins from the shared set by hand as frequently observed contaminants, which concedes that a rank cutoff lets contaminants through. ProteomeXchange/PRIDE 2018
• The same paper that motivates the nucleolar reading supplies its null hypothesis: MLL-AF4 is reported predominantly nucleolar, so every abundant nucleolar protein becomes a high-probability co-purification partner on compartment grounds alone. And the prediction that reading makes fails -- if nucleolar residency of the bait drove recovery, the index fusion should pull MORE of the block than the other six baits, and it pulls the FEWEST (7 of 26, tied last with MLL-GAS7). The report is also an ectopic-overexpression system, which this project treats as weak evidence about an endogenous protein. Siemund AL et al. 2022

HUMAN AML, myeloid, 2016 Clinical Cancer Research. This is NOT the index disease: AML is myeloid and adult-weighted, KMT2A::AFF1 disease is infant B-lineage, and NO MLL or KMT2A fusion appears anywhere in this paper. Read from the abstract only; the mechanistic detail (PCNA ubiquitination, in vitro ubiquitination assays, colony-forming assays) is taken from the abstract's own Experimental Design and not from the figures. n = 54 patients is a western-blot expression survey, not a genotyped cohort, and no MLL-rearrangement status is reported for those patients so it cannot be asked whether the KMT2A-rearranged subset behaves differently. Cited for one thing: PA2G4 is the only member of this nucleolar block with a published functional result in human leukaemia cells.

• The same paper's own framing is that Ebp1 has been reported with BOTH oncogenic and tumour-suppressor activities in different cancers, so a pro-proliferative reading in AML is context-specific and does not generalise. More sharply for this graph: an AML result cannot be transported to infant B-lineage KMT2A::AFF1, and the graph's measured base rate is that 21% of its load-bearing evidence carries a myeloid marker against 18% B-lineage - importing myeloid results is this project's commonest documented synthesis failure. Nguyen le XT et al. 2016
• Whatever Ebp1 does in AML, the route by which it reaches this graph is a co-purification from a T-ALL nuclear lysate with a tagged transgene, and the block it belongs to is the single most likely candidate in the whole table for an abundance artefact. The complex analysis in this repo was built to say so: in the index fusion's own column the CCT/TRiC chaperonin, a canonical AP-MS contaminant, is better represented (8 of 9 subunits, 6 at p<0.05) than the nucleolar block (7 of 26, 4 at p<0.05). ProteomeXchange/PRIDE 2018

Database record fetched live 2026-09-21. The subcellular-location comments carry ECO:0000269 experimental codes pointing at PubMed 16832058 and 19037095, which this beat did NOT open - so per this project's rule 91, the evidence code establishes that a curator read an experiment, not that those papers are the founding or best observation for the fact. The nucleolar assignment here is compartment-level only: UniProt gives NO fibrillar-centre, dense-fibrillar-component or granular-component sub-assignment for PA2G4, and this beat found none in the literature it read. That sub-nucleolar position is a BOUNDED ABSENCE, not a gap in the write-up.

• A protein whose nucleolar residency is switched by a PKC phosphosite is one whose compartment depends on the signalling state of the cell it was lysed from. Jurkat cells carrying a doxycycline-induced fusion transgene are not in a defined signalling state, so the fraction of PA2G4 that was nucleolar at the moment of lysis is unknown, and a co-purification cannot distinguish the nucleolar pool from the cytoplasmic one after a nuclear-lysate preparation. Skucha A et al. 2018

A counted negative with a date, the exact query, the querytranslation and the tool. Bounded by TITLES AND ABSTRACTS ONLY - supporting figures were not searched, and this project's rule 80 records a case where an H3K79me2 ChIP existed in a figure of a paper whose abstract never named the mark. Bounded also by the NAME LIST in data.bounded_literature.names_tried. The alias EBP1 is a short token and was title-checked: it returns nothing in this intersection at all, so there was nothing to screen. No archive query was run for PA2G4 specifically beyond PXD009338, which is the deposition of the very experiment in E1.

• An absence of publication is not an absence of measurement. PA2G4 is one of the 128 conserved interactors that WERE shRNA-screened in MLL-AF9 MOLM-13 in the source paper itself, with six shRNAs, so a functional measurement against an MLL fusion exists and was simply never tabulated - the per-gene score is measured-not-reported, not never-measured, and an afternoon with the authors' source data would close it. Skucha A et al. 2018