Selective in the disease and quiet everywhere else: MYB comes out cleanest
also called selectivity vs normal tissue · target triage · quiet and selective
A good drug target has to be needed by the cancer and not needed by the rest of the child. We had one list for each and had never put them side by side. Doing so puts MYB at the top.
The same thing, for a scientist
Crossing the cached **17,728-gene** KMT2A-rearranged selectivity scan with GTEx v8 across 54 tissues, after dropping **125 pan-essential** genes: 146 of the 150 most selective resolve, and **35 have a maximum normal-tissue median below 25 TPM**. **FLT3 ranks first and BCL2 fifth** — both already established handles — which is the calibration. The new result is **MYB**: selectivity −0.475 with a maximum normal-tissue median of **2.4 TPM** and a whole-body median of **0.3**, the cleanest profile in the set, and a gene this project's own ledger lists as unresolved.
The list
| gene | selectivity | max normal TPM | breadth (median of 54) | peak tissue |
|---|---|---|---|---|
| FLT3 | -0.843 | 22.5 | 0.4 | Brain_Cerebellar_Hemisphere (23) |
| MYB | -0.475 | 2.4 | 0.3 | Cells_EBV-transformed_lymphocytes (17) |
| FANCF | -0.440 | 12.7 | 4.2 | Cells_EBV-transformed_lymphocytes (13) |
| ZFP64 | -0.410 | 6.4 | 3.7 | Cells_EBV-transformed_lymphocytes (11) |
| BCL2 | -0.396 | 21.3 | 4.7 | Cells_EBV-transformed_lymphocytes (50) |
| BRIP1 | -0.346 | 7.3 | 0.2 | Cells_EBV-transformed_lymphocytes (9) |
| GFI1 | -0.312 | 9.7 | 0.5 | Cells_EBV-transformed_lymphocytes (13) |
| BCL11A | -0.292 | 10.8 | 1.1 | Skin_Not_Sun_Exposed_Suprapubic (15) |
| KAT6A | -0.285 | 23.4 | 15.2 | Lung (29) |
| TERT | -0.292 | 2.7 | 0.0 | Testis (3) |
Selectivity is the difference in mean DepMap gene effect between 11 KMT2A-rearranged and 43 other leukemia lines; more negative means more required in the rearranged lines. Normal exposure is the maximum median TPM across brain, heart, blood, liver/kidney and endocrine tissue.
The controls, and then the new part
FLT3 first, BCL2 fifth. Both are established handles in this disease and both are already in this graph. An unbiased scan crossed with an expression atlas recovering them unprompted is what makes the rest of the list worth reading.
MYB is the finding. Selectivity −0.475, maximum normal-tissue median 2.4 TPM, whole-body median 0.3 — essentially absent from adult tissue outside lymphoblastoid lines. The coverage ledger's standing weak spot list says "MEIS1 and MYB remain unresolved across designs". On this triage MYB is the single cleanest candidate in the graph, and it has been sitting unresolved.
The Fanconi pathway arrives again by a different road. FANCF (−0.440) and BRIP1 (−0.346, FANCJ) both make the top ten. Earlier cycles established the FA signal in KMT2A-rearranged lines as real and KMT2A-associated while excluding both rival mechanisms — aldehyde detoxification and replication stress. A genome-wide selectivity scan is a wholly different instrument from a pathway-level correlation, and it lands on the same pathway.
KAT6A is MOZ, the fusion that last cycle's provenance sweep found makes up half the deposition behind this graph's KDM4C claims; it appears here at −0.285.
BCL11A, KAT6A and TERT have no entry in this graph.
- This is a ranked question list, not a target list. Nothing here has been tested in an infant cell, an infant sample, or any KMT2A::AFF1 material beyond the 11-line panel.
- A favourable triage profile is not druggability. MYB is a transcription factor with no described ligandable pocket, and the analysis does not measure tractability at all.
- GTEx is blind to the tissue that matters most here. MYB's low blood TPM is compatible with an absolute requirement in haematopoietic stem and progenitor cells, which an adult whole-blood median cannot see.
- Do not drop a gene because it scores high on normal exposure. The same table's FLT3 row sits at 22.5 TPM and FLT3 inhibitors are used in this disease.
- Write entries for BCL11A, KAT6A and TERT. Three of the top ten selective-and-quiet genes have no entry in this graph, and KAT6A is MOZ - the fusion that makes up half the deposition behind the KDM4C claims.
- Run a proper pathway enrichment on the 146 scored genes. Two FA genes in a top ten is an observation; whether the FA pathway is enriched among selective-and-quiet genes is a test nobody has run.
- Re-run the selectivity axis when the index-fusion-models beat reports. If infant-derived material exists and is not in the 11-line panel, every number in this table is provisional.
- Add a druggability axis. Selectivity and normal-tissue exposure are two of three; ChEMBL activity counts and structural coverage are already fetched elsewhere in this project and would complete the triage.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.analysis
Crossing the 17,728-gene KMT2A-rearranged selectivity scan with GTEx v8 across 54 tissues, after excluding 125 pan-essential genes, 146 of the 150 most selective genes resolve and 35 have a maximum normal-tissue median below 25 TPM. FLT3 ranks first on selectivity (-0.843) and BCL2 fifth (-0.396); both are established therapeutic handles in this disease and were not supplied to the analysis.
Selectivity against normal-tissue exposure: crossing the 17,728-gene KMT2A-rearranged dependency scan with GTEx · Oncofusion knowledge graph (internal analysis) 2026What this evidence cannot show
Run 2026-09-20. Selectivity values are read from the cached scan and not recomputed, so they inherit its release and its 11-line panel. The pan-essential cut at mean effect <= -0.50 and the quiet cut at 25 TPM are both chosen to make the list readable, not calibrated against any outcome. Four of 150 symbols did not resolve in GTEx.
Evidence that argues against this (2)
- Recovering FLT3 and BCL2 is weaker calibration than it looks: both are among the most-studied genes in this disease precisely because they have large, reproducible dependency signals, so any method with any power recovers them. The controls show the analysis is not broken; they do not show it can discriminate among the less-obvious 33. Oncofusion knowledge graph (internal analysis) 2026
- The peak tissue for most of these genes is Cells_EBV-transformed_lymphocytes, which is a cell line and not an organ. That is exactly where a lymphoid dependency should appear and also exactly where an artefact of lymphoblastoid biology would, and this analysis cannot tell the two apart. GTEx Consortium / Broad Institute 2026
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.analysis
MYB has the cleanest combined profile in the set: selectivity -0.475 in KMT2A-rearranged lines, a maximum normal-tissue median of 2.4 TPM across brain, heart, blood, liver/kidney and endocrine tissue, and a whole-body median of 0.3 TPM across all 54 GTEx tissues, peaking in EBV-transformed lymphocytes at 17 TPM.
Selectivity against normal-tissue exposure: crossing the 17,728-gene KMT2A-rearranged dependency scan with GTEx · Oncofusion knowledge graph (internal analysis) 2026What this evidence cannot show
Selectivity from the cached 11-line panel; expression from adult post-mortem tissue. MYB is a transcription factor with no ligandable pocket described, so a favourable triage profile does not imply a tractable target - this graph's standing rule that a database fact is opposable on scope applies, and druggability is a separate axis this analysis does not measure.
Evidence that argues against this (2)
- MYB is required for normal haematopoiesis, which GTEx cannot see because whole blood is dominated by mature cells that no longer need it. A low blood TPM in an adult atlas is entirely compatible with an absolute requirement in the haematopoietic stem and progenitor compartment - which is the tissue an infant needs most and the one this analysis is blindest to. GTEx Consortium / Broad Institute 2026
- The ledger lists MYB as unresolved *across designs*, meaning previous analyses disagreed about it rather than never testing it. A new instrument agreeing with one side of an existing disagreement is not a resolution, and this entry should not be read as settling it. Oncofusion knowledge graph (internal analysis) 2026
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.analysis
Two Fanconi-anaemia pathway genes appear independently in the top ten selective-and-quiet hits: FANCF at -0.440 and BRIP1 (FANCJ) at -0.346. This is the FA signal arriving from a genome-wide selectivity scan, an instrument independent of the pathway-level correlation analyses that first found it.
Selectivity against normal-tissue exposure: crossing the 17,728-gene KMT2A-rearranged dependency scan with GTEx · Oncofusion knowledge graph (internal analysis) 2026What this evidence cannot show
Two genes from a pathway of more than twenty appearing in a top-ten list drawn from 146 is suggestive and is not a pathway enrichment test, which was not run. The earlier FA work established the signal was carried by myeloid lines with the lymphoid arm non-significant - the wrong lineage for this project - and nothing here revisits that.
Evidence that argues against this (2)
- Independence is partial at best. Both this analysis and the earlier FA work draw on the same DepMap Chronos matrix and the same 11-line KMT2A-rearranged panel, so a panel-level artefact would produce the signal in both. The instruments differ; the data do not. Oncofusion knowledge graph (internal analysis) 2026
- The FA hypothesis has already survived two rival mechanisms being excluded and still has no mechanism of its own, and its strongest correlate was MEN1 in myeloid lines. A third arrival without a mechanism raises the confidence that something is there without moving the question of what. Oncofusion knowledge graph (internal analysis) 2026