Oncofusion

start / graph / ana-dot1l-in-the-pax5-proximity-interactome

Analysisunrated confidence 0 of 5: unrated

DOT1L sits in a B-lineage proximity map, and we nearly did not look

also called PAX5 PXI DOT1L · PAX5 BioID 239 · B-lineage proximity DOT1L

The enzyme at the centre of this project's story turns up next to the master switch of B-cell identity, in the right kind of leukaemia cell. The result was sitting in a spreadsheet attached to a paper we had set aside.

The same thing, for a scientist

A BioID map of **PAX5** in pre-B cells identifies **239 proximity interactors**. Parsing S1 Table directly: **DOT1L is in the list**, and it is the **only** SEC or KMT2A-complex member there — no KMT2A, menin, AFF1, AFF4, ELL/ELL2, MLLT1/3/10, CDK9 or WRAD subunit. **BCL11A is also present.** This matters because this project's H3K79/DOT1L mechanism is overwhelmingly imported from murine myeloid MLL-AF9 systems, and this is DOT1L in a **B-lineage** proximity map.

Go deeper
How this was found, and the discipline it needed

The Q20 proximity beat rejected this paper — PAX5 is outside its protein set and IntAct serves no AFF1/AFF4/KMT2A row from it — and then flagged its own doubt: "I did not check its 239-protein supplementary list for any SEC or KMT2A-complex member. That is the first thing a repeat of this beat should do."

Doing it took three steps and the middle one nearly produced a wrong answer.

  1. Full text: 173,794 characters, <body> present, and zero occurrences of any SEC or KMT2A-complex symbol. On the body text alone the rejection looked correct.
  2. All nine supplementary spreadsheets, parsed as zip archives: hits for MEN1, CDK9, CCNT1, ELL2, DOT1L, ASH2L, RBBP5, LEO1, CTR9, MLLT1 and MLLT3. This list is almost entirely noise. S2–S9 are ChIP-seq target-gene tables and GO analyses — genome-wide lists running to 29,000 strings, in which a single occurrence of MEN1 means nothing at all. Reporting that list as 'the PAX5 interactome contains menin' would have been a clean, specific, wrong answer.
  3. The right table only. The captions identify S1 Table as "Identification of PAX5 proximity interactors (PXIs) in Pre-B cells"; the file numbering maps it to s002.xlsx. That sheet contains exactly 239 symbol-like entries, matching the paper's stated count, and DOT1L sits in it alphabetically between DNM1L and DPF2. One SEC/KMT2A-complex member, not eleven.
Why DOT1L in a B-lineage map is worth having

The provenance audits found that 54% of this graph's load-bearing mechanism comes from a non-index fusion, and the unstated-systems sweep found eight of sixteen resolved systems were murine MLL-AF9 myeloid. The H3K79/DOT1L story is the worst affected: the keystone CRISPR-ChIP partition is murine MLL-AF9, and mechanism-dot1l-prc1-antagonism spans myeloid, erythroid and pluripotent cells with zero B-lymphoid samples.

This is DOT1L recovered by proximity to PAX5, the master regulator of B-cell identity, in pre-B leukaemia cells. It does not make DOT1L a PAX5 binding partner and it says nothing about KMT2A::AFF1. What it does is put DOT1L in the right lineage, in a modern method, in a dataset this project had set aside.

BCL11A is in the same 239. This graph gained a gene-bcl11a entry in the same cycle, on entirely separate grounds — a selectivity-versus-normal-tissue triage. Two independent routes to the same protein in one cycle is worth noting and is not evidence of anything on its own.

Still open
  1. Is DOT1L a PAX5 target gene as well as a proximity interactor? The S2-S9 ChIP-seq tables were set aside as noise for the interactome question but they answer a different one.
  2. Does PAX5 proximity to DOT1L survive in a KMT2A-rearranged B-ALL line? The study's cells are murine pre-B and mouse B-ALL; the index system is human and fusion-driven.
  3. The Q20 beat's rejection was correct on its own terms and incomplete. Worth re-reading its other rejections the same way - it named this one itself, but the others were not flagged.
Every source behind this entry (2)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    S1 Table of the PAX5 BioID study contains exactly 239 symbol-like entries, matching the paper's stated count of 239 proteins, and DOT1L is among them. DOT1L is the only SEC or KMT2A-complex member in that table: KMT2A, MEN1, AFF1, AFF4, ELL, ELL2, MLLT1, MLLT3, MLLT10, CDK9, CCNT1, WDR5, RBBP5, ASH2L, DPY30, PAF1, CTR9, LEO1, CDC73 and PSIP1 are absent from it. BCL11A is present.

    What this evidence cannot show

    Retrieved and parsed by the coordinator on 2026-09-20: the Europe PMC supplementaryFiles archive for PMC6695195, then `pgen.1008280.s002.xlsx` opened as a zip and its sharedStrings read directly. **The mapping from 'S1 Table' to `s002.xlsx` is inferred from file ordering** (nine tables, nine xlsx files numbered s002–s010) rather than from an explicit statement, and the symbol count matching 239 is the check that it is right. The system is **murine pre-B / mouse B-ALL cells**, and the bait is PAX5, not KMT2A.

    Evidence that argues against this (2)
    • **Proximity is not binding, and this project has a concept entry saying so.** A BioID hit means within roughly 10 nm at some point, not in a complex. Nuclear proteins are labelled promiscuously and abundance drives recovery, so DOT1L's presence in a 239-protein nuclear proximity list is weak evidence of any specific relationship with PAX5. Oncofusion knowledge graph (internal analysis) 2026
    • The bait is PAX5 and the question this graph cares about is KMT2A::AFF1. Nothing here involves a KMT2A fusion, and PAX5 is not known to be part of the SEC. The value is lineage context for DOT1L, not a new interaction in the index system, and reading it as the latter would repeat the import error the provenance audits measured. PLoS Genetics 2019
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Searching all nine supplementary spreadsheets indiscriminately returns apparent hits for MEN1, CDK9, CCNT1, ELL2, DOT1L, ASH2L, RBBP5, LEO1, CTR9, MLLT1 and MLLT3, but S2 to S9 are ChIP-seq target-gene tables and GO analyses containing up to 29,000 strings, in which a single occurrence carries no information. Restricting to the proximity-interactor table leaves one member, DOT1L.

    What this evidence cannot show

    Counted by the coordinator on 2026-09-20. The point is methodological rather than biological: it records how a plausible eleven-member result was avoided, not a finding about PAX5.

    Evidence that argues against this (1)
    • Dismissing S2-S9 wholesale may discard something real. Those tables are PAX5, RUNX1, IKZF1 and FLI1 ChIP-seq target genes, and a KMT2A-complex member appearing as a transcriptional TARGET of the B-lineage network would be a different and possibly interesting fact from appearing as a proximity interactor. Nobody has looked at it that way, here or elsewhere. PLoS Genetics 2019
Connected to