Oncofusion

start / graph / gene-ptbp1

Genemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q6-interactome-rna-layer

PTBP1 — the strongest RNA-layer dependency in KMT2A-rearranged lines, found by a domain-focused screen and missed by this beat's own gene panel

also called PTB · hnRNP I · HNRNPI · PTBP · polypyrimidine tract binding protein 1

PTBP1 is a splicing protein that KMT2A-rearranged leukaemia cells need much more than other leukaemia cells do — the largest such effect anywhere in the RNA machinery. It was found by a screen that cut proteins by their functional parts rather than by gene name, which is why a gene-by-gene search had missed it.

The same thing, for a scientist

A domain-focused CRISPR screen of 3,182 sgRNAs against 527 RNA-enzymatic and RNA-binding domains in 341 RNA-associated proteins identified PTBP1 and the m6A reader RBM15 as AML dependencies **with a bias toward the KMT2A-rearranged subtype**; all four PTBP1 RNA-binding domains are required for KMT2A-r AML proliferation, PTBP1 is not required for myelopoiesis, and its suppression causes cell-cycle arrest, apoptosis and myeloid differentiation, disrupting the KMT2A-r-essential transcriptional programme and splicing, with CLIP-seq showing preferential binding to IKZF1, MEF2C, EZH2, SIK3 and PBX3 [E1]. This project's independent DepMap scan reproduces the bias: PTBP1 is **-0.247 (B-lymphoid) / -0.382 (myeloid)** against 43 other leukaemia lines, the **0.0th percentile** of a 466-gene random null, concordant in both lineage halves and not floor-compressed at 50.7% of lines depleted [E2]. It is the only RNA-layer gene in a 79-gene panel to reach that percentile.

Go deeper
How this entry was found, and the methodological lesson

This beat ran three enumeration routes. PTBP1 came from the literature route and was absent from the dependency route's own panel, because that panel was built from m6A, splicing-factor, helicase and decay vocabulary and PTBP1 is none of those things in the way the vocabulary was written. It was added to the scan only after the paper was read, and it immediately outscored every gene that had been there from the start 2.

That is the beat's own version of the failure it was dispatched to correct. gene-igf2bp3 diagnosed enumeration-from-complex-membership; the panel here was enumeration-from-functional-category, and it has the same shape of blind spot: it finds the members of the categories you thought of.

The screen that found PTBP1 avoided this by not enumerating genes at all. It tiled sgRNAs across RNA-binding and enzymatic domains 1, so membership was defined by protein structure rather than by a curator's category. Any future enumeration in this graph should note that a domain-tiled screen and a gene-list scan have different coverage, and that the gene list is the narrower of the two.

Why this is the RNA-layer candidate worth taking seriously

The rest of this beat's genes fail one or more of four tests. PTBP1 passes all four.

It is selective, not pan-essential. 50.7% of DepMap lines depleted, well clear of the 90% floor-compression threshold at which a difference stops being interpretable 2.

It is concordant across the lineage split. -0.247 lymphoid and -0.382 myeloid, the same sign in both halves. METTL3, MSI2, YTHDF2, IGF2BP1 and ALKBH5 are all lineage-discordant in the same scan, and for those the pooled number is a lineage average of contradictory results 2.

The effect size is real, not just the percentile. -0.33 pooled Chronos against a null of SD 0.075. This project corrected itself in cycle 45 for ranking narrowness instead of effect; PTBP1 is at the extreme on both.

There is a normal-tissue argument. The screen reports PTBP1 is not required for myelopoiesis while KMT2A-r AML needs it 1 - the shape a window has.

It still falls far short of menin (-0.871 / -0.863) and it is an AML result 1. There is no B-lymphoid KMT2A::AFF1 experiment, and the DepMap lymphoid signal rests on four cell lines.

The targets, and the one that connects to the rest of the graph

PTBP1 CLIP-seq preferentially binds IKZF1, MEF2C, EZH2, SIK3 and PBX3 1. Three of those already have standing in this graph's account of KMT2A-rearranged disease - MEF2C as a KMT2A-r programme gene, IKZF1 as a B-lymphoid transcription factor with prognostic weight, PBX3 as a HOX cofactor.

That gives a testable prediction rather than a coincidence: if PTBP1's requirement runs through fine-tuning the expression and splicing of exactly the transcripts the fusion depends on, then PTBP1 loss and menin inhibition should converge on the same programme, and combining them should be sub-additive rather than synergistic. The IGF2BP3 case went the other way - depletion sensitised to menin inhibition 3 - so the prediction is discriminating.

Facts and figures
gene symbolPTBP1
rna functionpolypyrimidine-tract binding splicing regulator with four RNA-binding domains; controls processing and splicing of oncogenic transcripts
discovery routedomain-focused CRISPR screen of RNA-binding and enzymatic domains, not a gene-list enumeration
uniprotP26599
length AA557
uniprot resolutiongene_exact:PTBP1 AND organism_id:9606 AND reviewed:true; one hit; primary gene name read back as PTBP1; length 557 aa checked. The alias hnRNP I collides with hnRNP nomenclature and was not used as a search term.
subnuclear localizationThe PERINUCLEOLAR COMPARTMENT - PTB defines it and is required for its integrity (PMID 12808040) - on a diffuse nucleoplasmic background. The PNC is a transformation-associated structure, so this address is a property of the malignant state.
shown for fusionDEPENDENCY YES in KMT2A-rearranged AML; LOCALISATION NO in any fusion. PTBP1 is a validated dependency biased to the KMT2A-rearranged AML subtype (PMID 42637878), which is a fitness measurement. Nobody has imaged PTBP1, or scored PNC prevalence, in any KMT2A-rearranged cell: the bounded localisation query returned 0.
essentiality
{
  "depmap_common_essential": false,
  "chronos_mean": -0.5247,
  "pct_lines_depleted": 50.7,
  "n_lines": 1208,
  "kmt2a_r_lymphoid_diff": -0.2465,
  "kmt2a_r_myeloid_diff": -0.3815,
  "z_lymphoid": -0.782,
  "z_myeloid": -1.211,
  "null_percentile": 0,
  "floor_compressed": false,
  "lineage_discordant": false
}
clip targets
[
  "IKZF1",
  "MEF2C",
  "EZH2",
  "SIK3",
  "PBX3"
]
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": -0.8753,
  "chronos_mean_other_leukaemia": -0.5429,
  "difference": -0.3324,
  "pan_essentiality_all_lines": -0.5247,
  "selectivity_rank": 66,
  "rank_denominator": 17728,
  "provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
  "how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}
localization
{
  "subcellular": "Predominantly NUCLEAR with a real cytoplasmic pool, including P-bodies. UniProt P26599: Nucleus, Cytoplasm and Cytoplasm/P-body, all three on ECO:0000269 PubMed:40965981 - a 2025 paper on PTBP1 variants with altered nucleocytoplasmic distribution, so the coded citation here is unusually on-topic. HPA immunofluorescence gives a single main location, Nucleoplasm, reliability Supported.",
  "subnuclear": "THE PERINUCLEOLAR COMPARTMENT (PNC). PTB is one of the RNA-binding proteins that DEFINES the PNC, a nuclear substructure found at the nucleolar periphery; siRNA depletion of PTB reduces both the number of PNC-containing cells and PNC size, so PTB is not merely resident but required for the compartment's integrity (PMID 12808040). HPA's nucleoplasm call does not resolve the PNC, and there is no contradiction - the PNC is a small perinucleolar structure.",
  "how_shown": "immunofluorescence of the PNC plus microinjection of the pol III inhibitor Tagetin and siRNA depletion of PTB (PMID 12808040); nucleocytoplasmic distribution of variants (PMID 40965981)",
  "tissue_of_evidence": "HeLa and other transformed human lines for the PNC work. NOTHING HAEMATOPOIETIC has been imaged, which is striking given that PTBP1 is this graph's strongest RNA-layer dependency in KMT2A-rearranged AML.",
  "the_compartment_is_a_state_marker": "The PNC is present in TRANSFORMED cells and is used as a prevalence marker of malignancy - it has been scored as an independent prognostic factor in paediatric Ewing sarcoma and used as the readout of an anti-metastatic compound screen. So for PTBP1 the subnuclear answer and the cell-state answer are the same fact.",
  "node": "localization-paf1-nucleoplasmic-puncta-versus-a-nucleolar-pool"
}
cell type
{
  "haematopoietic_answer": "ACUTE MYELOID LEUKAEMIA BLASTS, with a KMT2A-REARRANGED BIAS, and specifically NOT normal myelopoiesis. A domain-focused CRISPR screen found PTBP1 a dependency in AML biased to the KMT2A-rearranged subtype; all four RNA-binding domains are required for KMT2A-r AML proliferation, and PTBP1 is not required for myelopoiesis (PMID 42637878, already cited in this entry as E1).",
  "why_the_gap_is_notable": "PTBP1's cell-type answer is the most fusion-specific in this tranche and its localisation answer is the least - the compartment work is entirely in HeLa-class lines and nobody has looked at the PNC in a KMT2A-rearranged blast.",
  "expression_atlas": "HPA Blood Atlas: low immune cell specificity, Detected in ALL immune cell types; RNA tissue specificity low."
}
cell state
{
  "state": "THE TRANSFORMED STATE ITSELF, read through the PNC. The perinucleolar compartment forms in transformed cells, its prevalence tracks malignancy, and PTB is required to build it (PMID 12808040). Separately, PTBP1 suppression in AML causes cell-cycle arrest while sparing normal myelopoiesis (PMID 42637878).",
  "why_this_is_a_real_state_and_not_a_label": "PNC prevalence is scored cell-by-cell and has been used as an independent prognostic variable and as a drug-screen endpoint, so it behaves as a measurable state rather than as a description."
}
localization bounded absence
{
  "date": "2026-09-21",
  "tool": "NCBI E-utilities esearch, db=pubmed",
  "shape": "gene clause AND localisation clause AND KMT2A/MLL-fusion clause, one [tiab] tag per term (brief rule 94), every cytogenetic term quoted (brief rule 95)",
  "querytranslation_fusion_clause": "\"KMT2A\"[Title/Abstract] OR \"MLL\"[Title/Abstract] OR \"MLL1\"[Title/Abstract] OR \"MLL-AF4\"[Title/Abstract] OR \"MLL-AF9\"[Title/Abstract] OR \"MLL-ENL\"[Title/Abstract] OR \"KMT2A-rearranged\"[Title/Abstract] OR \"MLL-rearranged\"[Title/Abstract] OR \"t 4 11\"[Title/Abstract] OR \"t 9 11\"[Title/Abstract] OR \"MLL fusion\"[Title/Abstract]",
  "count": 0,
  "controls": {
    "fusion_clause_alone": 8743,
    "gene_clause_alone": 11153
  },
  "reading": "Counted in TITLES AND ABSTRACTS only; supporting figures were not searched (brief rule 80)."
}
How this could bite us back
  • Depleted in half of all DepMap lines. This is a broad dependency with a KMT2A-rearranged bias on top, and the therapeutic-window argument rests on one normal-myelopoiesis readout.
  • Splicing-factor inhibition has an unhappy clinical history in myeloid disease; the class carries on-target haematologic toxicity that shows up in patients and not in lines.
  • All four RNA-binding domains being required is a reason the protein is hard to drug, not a reason it is easy: there is no single catalytic pocket to occupy.
  • If PTBP1 works by fine-tuning the same transcripts the fusion drives, a cell that loses PTBP1 can in principle restore them by changing their transcription, which is the axis menin inhibitors already contest.
Still open
  1. Is PTBP1 required in KMT2A::AFF1 B-lymphoid cells? The screen is entirely AML and the DepMap lymphoid signal rests on four lines.
  2. Does PTBP1 loss combine sub-additively with menin inhibition, as the shared-programme model predicts, or synergistically as IGF2BP3 depletion does?
  3. Are the PTBP1-bound transcripts IKZF1, MEF2C and PBX3 mis-spliced in KMT2A-rearranged disease, or only differently abundant?
  4. Do PTBP2 and PTBP3, both flat in this scan, compensate on PTBP1 loss in any lineage?
  5. PTBP1 is required to build the perinucleolar compartment (PMID 12808040) and is a KMT2A-rearranged-biased AML dependency (PMID 42637878), and nobody has connected the two. Do KMT2A-rearranged blasts carry PNCs, does PNC prevalence track the KMT2A-r bias of the dependency, and does the PNC dissolve when PTBP1's RNA-binding domains are disabled in those cells? One immunofluorescence panel would test whether the dependency is a compartment dependency.
Every source behind this entry (9)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    A domain-focused CRISPR screen using 3,182 sgRNAs targeting 527 RNA-enzymatic and RNA-binding domains across 341 RNA-associated proteins identified PTBP1 and the m6A reader RBM15 as dependencies in AML with a bias toward the KMT2A-rearranged subtype; all four PTBP1 RNA-binding domains are required for KMT2A-r AML proliferation, PTBP1 is not required for myelopoiesis, its suppression causes cell-cycle arrest, apoptosis and myeloid differentiation, and CLIP-seq shows preferential binding to IKZF1, MEF2C, EZH2, SIK3 and PBX3 transcripts.

    PMID 42637878DOI
    What this evidence cannot show

    Abstract read; full text not read. AML throughout - the screen and validation are myeloid, with no B-lymphoid or KMT2A::AFF1 arm. 'Bias toward the KMT2A-rearranged subtype' is the abstract's phrasing and the magnitude of that bias is not quantified in the abstract.

    Evidence that argues against this (2)
    • PTBP1 is depleted in 50.7% of all 1,208 DepMap lines and has a Chronos mean of -0.525 across every lineage. A gene required in half of all cancer lines is substantially a general dependency, and the KMT2A-r bias sits on top of a large lineage-independent requirement rather than replacing it. Oncofusion knowledge graph (internal analysis) 2026
    • 'Not required for myelopoiesis' is a normal-tissue claim measured in a specific assay. This graph's menin work found preserved steady-state progenitors with a long-term HSC defect that appeared only under competitive transplantation, so an unstressed normal-haematopoiesis readout is the condition least likely to find the toxicity that matters. Oncofusion knowledge graph (internal analysis) 2026
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.analysis

    DepMap CRISPR (Chronos) gives PTBP1 a mean gene effect of -0.525 across 1,208 lines with 50.7% depleted below -0.5, and in the KMT2A-rearranged index arm a difference of -0.247 in four B-lymphoid lines and -0.382 in seven myeloid lines against 43 other leukaemia lines - the 0.0th percentile of a 466-gene random null (mean -0.005, SD 0.075), concordant in sign across the lineage split and not floor-compressed. It is the strongest of 74 RNA-layer genes tested and the only one to reach the 0.0th percentile.

    What this evidence cannot show

    Four B-lymphoid index lines; the lymphoid number alone would not be a result. PTBP1 was added to the panel only after the literature route surfaced it, so its ranking is not the output of a blind scan and should not be quoted as if it were.

    Evidence that argues against this (2)
    • MEN1 is -0.871 / -0.863 in the same arm, roughly 2.8-fold larger, and KMT2A itself is -0.436 / -0.601. PTBP1 is the best RNA-layer gene and is still well short of the fusion's own chromatin dependencies, so it is a candidate for combination rather than for replacement. Oncofusion knowledge graph (internal analysis) 2026
    • A gene added to a panel after reading the paper that nominates it will rank well by construction if the paper is right. The independent part of this result is the lineage-split agreement in a different dataset from the screen's, not the ranking. Cao Z et al. 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Depletion of Igf2bp3 sensitises MLL-AF4 leukaemia to menin-MLL inhibition on cell growth and leukaemic initiating cells, both manipulations increase differentiation, and IGF2BP3 knockdown had a greater effect on survival and disease attenuation than the menin-MLL inhibitor MI-503 alone.

    PMID 38048400DOI
    What this evidence cannot show

    Abstract read; full text not read. Murine MLL-AF4; genetic depletion versus a tool-compound menin inhibitor, which is not a like-for-like comparison. Cited here to set up a discriminating prediction about PTBP1, not as evidence about PTBP1.

    Evidence that argues against this (1)
    • No combination experiment with PTBP1 and a menin inhibitor has been reported, so the sub-additivity prediction in this entry is untested and should be read as a proposal. Cao Z et al. 2026
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.analysis

    PTBP1 was absent from this beat's original 76-gene RNA-metabolism panel and was added only after the literature route surfaced the domain-focused screen; on addition it ranked first of all RNA-layer genes.

    What this evidence cannot show

    A statement about this project's own process, verifiable from the panel definition in run.py and the comment recording why PTBP1 was added.

    Evidence that argues against this (1)
    • Recording a miss is not the same as fixing the method. The panel is still a gene list and still has the same class of blind spot for whatever the next paper turns up; only a domain-tiled or genome-scale design removes it. Cao Z et al. 2026
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProtKB P26599 (PTBP1_HUMAN) is human PTBP1, 557 amino acids; resolved by gene_exact query with the primary gene name read back as PTBP1 and the length checked.

    What this evidence cannot show

    Single reviewed human hit with matching primary gene name. PTBP1's alias hnRNP I collides with hnRNP nomenclature and was deliberately not used as a PubMed search term.

    Evidence that argues against this (1)
    • The accession covers one protein; the screen reports all four of its RNA-binding domains as required, so the druggable unit is the whole protein rather than a pocket, and the accession says nothing about whether any of the four domains is ligandable. Cao Z et al. 2026
  6. E-mll-enl-direct-target-metabolicstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    PTBP1 is a direct MLL-ENL target gene whose knockdown is rate-limiting for MLL-ENL-induced leukaemia and reduces glucose uptake, lactate production and the PKM2 isoform, placing a metabolic phenotype directly downstream of a gene the fusion binds.

    Exploitable metabolic dependencies in MLL-ENL-induced leukemia · Garcia-Cuellar MP, ... Slany RK · Blood Advances 2020
    PMID 32766858DOI
    What this evidence cannot show

    Abstract read on 2026-09-20; full text not read. **Strong within the reported system, which is MLL-ENL** - a myeloid fusion, not the index infant B-lineage KMT2A::AFF1. This is the best-SPECIFIED causal chain in this graph from the fusion to a metabolic phenotype, which is a statement about specification rather than about strength: 'direct target gene' plus a knockdown phenotype is not the same as showing the metabolic change is what makes the leukaemia grow.

    Evidence that argues against this (1)
    • PKM2 is an ISOFORM of PKM, not a gene, so an isoform-level readout cannot be verified by accession and length the way this graph checks gene claims - and `gene_exact:PKM2` silently resolves to the parent record. The metabolic readouts here are also the standard Warburg panel, which shifts with proliferation rate in almost any system and is weak evidence of a specific rewiring. Neville D et al. 2026
  7. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    PTB is one of the RNA-binding proteins that defines the perinucleolar compartment, and siRNA depletion of PTB reduces both the number of PNC-containing cells and the size of the compartment, so PTB is required for PNC integrity; the PNC is present in transformed cells and depends on RNA polymerase III transcripts.

    PMID 12808040DOI
    What this evidence cannot show

    HeLa cells, 2003, non-haematopoietic. The paper studies PTB by that name; PTBP1 is the current symbol and PTB[tiab] is a contaminated token (it also abbreviates pulmonary tuberculosis and phosphotyrosine-binding), so hits for it must be title-checked. Abstract-level reading. PNC presence in a KMT2A-rearranged blast has never been reported either way.

    Evidence that argues against this (1)
    • All three UniProt compartments rest on ONE publication, and that publication is about disease VARIANTS with altered nucleocytoplasmic distribution - a study design that selects for the trafficking phenotype it reports. The wild-type ratio between nucleus, cytoplasm and P-bodies is a comparator inside a variant paper, not an independently established quantity, and HPA's single main 'Nucleoplasm' call does not corroborate the cytoplasmic or P-body pools at all. Two sources, and they agree on one compartment out of three. Human Protein Atlas 2026
  8. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    UniProt P26599 places PTBP1 in the nucleus, the cytoplasm and cytoplasmic P-bodies, all three on a single coded experimental citation (PubMed:40965981) about PTBP1 variants with altered nucleocytoplasmic distribution; HPA immunofluorescence gives a single main location, Nucleoplasm, at reliability Supported.

    What this evidence cannot show

    All three UniProt compartments rest on ONE publication, and that publication is about disease variants, so the wild-type distribution is a comparator inside it rather than its subject. Neither source resolves the perinucleolar compartment. Nothing haematopoietic.

    Evidence that argues against this (2)
    • This entry's cell-state answer - that the PNC marks the transformed state - is directly undercut by newer work: alpha-satellite RNA marks the perinucleolar compartment in NAIVE HUMAN EMBRYONIC STEM CELLS, which are not transformed. If the PNC is a property of a permissive nucleolar/chromatin configuration rather than of malignancy, then PNC prevalence is not a malignancy readout and the proposed experiment in KMT2A-rearranged blasts would need a naive-pluripotent-like control. PTB is also not the compartment's only defining component - CDK13 is one too - so 'PTB defines the PNC' overstates a shared dependency. Genes Dev 2026
    • Second, weaker objection to the same claim: the compartment is defined operationally by the proteins used to stain for it, so 'PTB is required for PNC integrity' partly restates that PTB is one of the markers by which the PNC is recognised. CDK13's identification as a component shows the marker set is still growing, which means the PNC's composition - and therefore what a PTB-depletion phenotype is measuring - was not settled when the 2003 experiment was done. PLoS One 2016
  9. E8moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    Human Protein Atlas v25.1 records PTBP1 with the single main subcellular location Nucleoplasm at reliability Supported and detects it in all immune cell types at low specificity.

    What this evidence cannot show

    Antibody immunofluorescence at a resolution that does not separate the perinucleolar compartment from general nucleoplasm, so this is consistent with the PNC literature rather than in tension with it. RNA for the blood-cell call.

    Evidence that argues against this (1)
    • This row is neutral rather than supporting, and citing it as agreement would be a mistake. HPA's imaging resolution and scoring vocabulary do not separate a small perinucleolar structure from general nucleoplasm - the PNC has historically required dedicated immunohistochemical protocols to detect and score. So a 'Nucleoplasm' main call is compatible with a PNC-positive cell and with a PNC-negative one, and it contributes nothing either way to this entry's subnuclear record. Methods Mol Biol 2008
Connected to