The leukaemia term hid the answer for DPY30 — a rule-77 demonstration with clean rule-67 controls
also called DPY30 bounded absence · rule 77 in haematology
Searching for DPY30 and any cell-state word inside the leukaemia literature returns nothing. Dropping the word leukaemia from the same search returns the paper that answers the question. The search was working perfectly and the conclusion would have been wrong.
The same thing, for a scientist
Bounded query on 2026-09-20: `(DPY30 OR "Dpy-30" OR "DPY-30" OR Saf19)[tiab] AND <18-term cell-state clause> AND <leukaemia clause>` returns **0**. Both operands are large — the gene clause alone 121 records, the state clause alone 96,743 — so brief rule 67's control passes and the instrument can demonstrably return a one. Removing ONLY the leukaemia clause returns **6**, and one of them is PMID 27647347, *Dpy30 is critical for maintaining the identity and function of adult hematopoietic stem cells* (J Exp Med 2016) — the single best cell-type-and-state result for this gene [E1][E2].
Why this is rule 77 and not rule 67 or rule 71
Rule 67 says never trust a zero from a query you have not shown can return a one. That control passed here. Rule 71 says a pathway is often named by its drug, so widen the vocabulary; the vocabulary was already eighteen terms wide. Rule 70 says bound a noncoding element by its name list; not applicable.
The failure is rule 77: the disease term. DPY30's cell-state biology is published in normal haematopoiesis, and scoping to leukaemia — which is the correct scope for almost every other question this project asks — removed it. Rule 77 was written from a PD-L1 example in pancreatic cancer; this is the same failure inside haematology, where the scoping feels even safer.
The generalisable practice
For any cell-state or cell-type question about a gene, run the query without the disease term as a matter of course, not as a rescue after a zero. Normal-cell state biology is where conditional-knockout work lives, and conditional-knockout work is the best class-2 and class-3 evidence available for the chromatin genes in this graph.
The same move was worth making for KAT8 (adult-versus-fetal haematopoiesis), for PAF1 (CDC73 HSC maintenance) and for MLLT3 (human HSC self-renewal). In all four cases the strongest cell-context evidence for a gene this graph studies as a leukaemia gene comes from a normal-haematopoiesis paper.
| status | run |
|---|---|
| result summary | Scoping a cell-state query to leukaemia returns 0 for DPY30 and hides the J Exp Med adult-HSC result. Rule 67 controls pass; the error is rule 77. |
inputs
[ "NCBI esearch db=pubmed, 2026-09-20" ]
queries — the numbers behind the chart
[
{
"label": "scoped to leukaemia",
"count": 0
},
{
"label": "identical query, leukaemia clause removed",
"count": 6
},
{
"label": "control: gene clause alone",
"count": 121
},
{
"label": "control: state clause alone",
"count": 96743
}
]genes where the same move paid
[ "DPY30", "KAT8", "PAF1", "MLLT3" ]
- A gene dismissed as unstudied because of a disease-scoped zero would be dropped from a target list on the strength of a search artefact.
- Conversely, importing a normal-haematopoiesis requirement into a leukaemia target rationale is the class-2 context error this beat's concept entry warns about; dropping the disease term widens recall and weakens specificity at the same time.
- How many of this graph's existing bounded absences carry a disease term, and how many of those would change if it were dropped?
- Is there a systematic way to flag a cell-state question as one that should be run unscoped?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
A PubMed query pairing DPY30 and its aliases with an eighteen-term cell-state clause and a leukaemia clause returns 0 records; the identical query without the leukaemia clause returns 6; the gene clause alone returns 121 and the state clause alone 96,743.
NCBI PubMed E-utilities esearch, db=pubmed, queried 2026-09-20 · NCBI Entrez Programming Utilities 2026What this evidence cannot show
Counts as of 2026-09-20 and they will drift. Bounded by titles and abstracts; supporting figures not searched. The zero is a true measurement of the scoped literature and a false answer to the question asked.
Evidence that argues against this (1)
- Dropping the disease term is not free. It widens recall and weakens specificity, and four of the six unscoped hits are in non-haematopoietic systems (goat testis, neural stem cells, human embryonic stem cells, HoxB4 lineage differentiation). The rescue that recovered the J Exp Med result would also have pulled in those, so the correct practice is to run both queries and read the difference, not to abandon the disease term. Open Targets Platform GraphQL API 2026
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Dpy30 is critical for maintaining the identity and function of adult haematopoietic stem cells in mice.
Dpy30 is critical for maintaining the identity and function of adult hematopoietic stem cells. · Yang Z, Shah K, et al. · The Journal of experimental medicine 2016What this evidence cannot show
Murine, normal haematopoiesis, not leukaemia and not human (brief rule 73). Abstract read in this beat; full text not opened. DPY30 acts as an ASH2L-specific stabiliser within the SET1/MLL complexes, so the phenotype is a complex-level phenotype read out on one subunit.
Evidence that argues against this (1)
- PUBLISHED CONFLICTING RESULT, and it strengthens this entry's rule-77 thesis while correcting the 'single best result' claim. Yang et al. place DPY30 in HUMAN CD34+ haematopoietic progenitors rather than in mouse: knockdown impairs their ex vivo proliferation and differentiation, and the effect is lineage-DIRECTIONAL rather than a general loss of identity - it impairs differentiation into the myelomonocytic lineage while POTENTLY PROMOTING haemoglobin production and altering erythroid differentiation kinetics. Morpholino knockdown in zebrafish gives severe defects in haematopoietic development, partially rescued by dpy30 mRNA. The same paper reports a crucial role for DPY30 in the growth of several MLL1-fusion leukaemia cell lines - i.e. the leukaemia-scoped query that returned zero was looking for content this paper contains. The murine adult-HSC 'identity and function' summary is one of at least two published cell-state answers and it is not the one in human cells. Yang Z et al. 2014
- part of →DPY30 — the smallest COMPASS subunit, and the only one that behaves like a selective dependency
the bounded cell-state measurement for this gene
- related to →ASH2L — the WRAD subunit that has DPY30's essentiality profile and none of its co-dependency signal
DPY30 acts as an ASH2L-specific stabiliser, so the phenotype is a complex-level phenotype read out on one subunit
- refutes →DPY30 — the smallest COMPASS subunit, and the only one that behaves like a selective dependency
the best cell-state result for DPY30 is not the murine adult-HSC one: knockdown in HUMAN CD34+ progenitors impairs myelomonocytic differentiation while POTENTLY PROMOTING haemoglobin production, with severe haematopoietic defects in zebrafish and a crucial role in MLL1-fusion leukaemia line growth (PMID 25139354) - which is content the leukaemia-scoped query that returned zero was looking for