Prioritisation: ranking targets, not drugs

Targets are ranked by a small, transparent score over three CITED tiers -- burden, unmet need, and mechanistic directness -- with DECLARED weights (B 0.40, U 0.35, D 0.25). The top target is the shared insulin node. The γ-|hsp| read is carried as structural context and is NEVER folded into the clinical score.

The analog of analgesic M10. Burden and unmet need lead the weights by editorial choice; the ranking is [F] from cited tiers plus declared weights, not a [V] engine output. The firewall forbids equating a promoter-stiffness read with a clinical magnitude, so γ sits beside the score, never inside it.

The ranking

targetleverBUDscoreγ-|hsp| (context)
INSRS15454.650.7040
MC4RS24444.000.5522
LEPRS14433.750.6758
UCP1S24433.750.6413
PPARGS14343.650.6309
PDK4S34423.500.6453
ADRB3S23333.000.6694
GHRLS23333.000.6071
TNFS33322.75

Declared weights, not tuned

score = 0.40·B + 0.35·U + 0.25·D over cited 1-5 tiers. The weights are an explicit editorial choice (need leads); they were not tuned to produce a desired ordering. Ranking targets -- not drugs -- keeps the output a research prioritisation, not a treatment list.

Firewall

score = w_B*B + w_U*U + w_D*D over CITED 1..5 tiers; the gamma-|h_sp| read is carried as structural context and is NEVER folded into the clinical priority score (a promoter-stiffness read is not a clinical magnitude). Ranking is [F] from cited tiers + declared weights.