The whole protein-coding genome: menin is still first, and a metabolic pathway appears (run)

The 4,589-gene scan left an obvious open question, recorded in its own entry: "Scanning the remaining ~14,000 protein-coding genes would cost roughly four hours of API time. Is anything above MEN1 hiding there?"

No. Across 17,728 genes — the full HGNC protein-coding set with DepMap data — MEN1 is the single most selectively required gene in KMT2A-rearranged leukemia lines, at the 0.006th percentile 1.

That is the strongest form this project's central premise can take from this dataset. It is also the form most exposed to being wrong: MEN1 competed against every measurable gene, unflagged, and a poor rank would have been unmissable.

The cycle-22 strikes hold at genome scale, having been reached without any knowledge of them: CDC73 11,645 (65.7th percentile), KDM4C 14,509, RBBP5 14,622, MLLT3 14,902, ASH2L 17,248, KDM6A 17,398 and CTR9 17,413 of 17,728 (98.2nd percentile) — every one with a positive difference, i.e. marginally less required in rearranged lines. DPY30 sits at 2,590 (14.6th), unremarkable for the sixth independent design.

JMJD1C at rank 23 is a useful check. It is a documented MLL-fusion dependency the scan was not told about, and its near-zero pan-essentiality (-0.012) means it is selective rather than generally required. Together with ZFP64 in the smaller scan, that is two known mechanisms recovered from dependency data alone.

Four genes in iron and haem handling rank in the top 47 of 17,728 — the top 0.27% — and none is pan-essential:

All four accessions were resolved by exact gene symbol and the primary gene name read back 2.

This contradicts the explanation this project adopted one day earlier. UROS appeared at rank 15 in the smaller scan; a literature search found nothing connecting it to leukemia, and the beat recorded a deliberate clean negative with a mechanistic explanation: UROS loss causes accumulation of porphyrin isomer I, the toxic-metabolite mechanism of congenital erythropoietic porphyria, which would harm any cell with upstream haem flux regardless of the fusion.

That was a careful, well-reasoned conclusion on the evidence available, and the pathway view breaks it. The gain-of-toxic-metabolite model makes a directional prediction: reduce flux into the pathway and the toxicity should fall. ALAS1 is the first and rate-limiting enzyme — knocking it out reduces flux, so under that model ALAS1 loss should be neutral or protective. Instead ALAS1 loss is selectively deleterious, and so is ALAD's at enzyme 2. Blocking the tap does not rescue; it hurts more.

The simpler reading is that KMT2A-rearranged lines have an elevated requirement for haem synthesis itself. This project's own graph already holds an unexplored thread in that direction: SETD1A controls pause release at haem genes. That was noted and set aside as pointing the wrong way. With four pathway members ranking, it deserves a proper look.

Eleven lines, none infant-derived. The panel is unchanged and remains the binding limitation on every conclusion here.

Same statistic as the specificity analysis. This is not methodologically independent of it; what is new is the absence of a prior and the complete ranking context.

Pan-essential contaminants persist. DNAJC9 at 4, TINF2 at 6, RPL13A at 19 and MCL1 at 18 are more depleted in rearranged lines because the panel is more sensitive, not because of the fusion. CDK6 at 13 sits on the boundary at -0.632 and should be read with that in mind even though its biology is supported elsewhere.

A rank is a lead, not a target. The scan-hits beat found that a rank buys about one afternoon of literature search, and that roughly half the time that search pays.