NULL: no promoter in the torpor panel marks hibernation
Widen the §11 question from one gene to the whole declared torpor program. Across 8 fuel-switch / BAT-identity genes read over 14 species (7 hibernators vs 5 non-hibernators among endotherms), the number of genes whose promoter γ cleanly separates hibernators from non-hibernators is 0. Every group-mean gap that exists co-signs and co-scales with the group GC gap (cross-gene Pearson r(Δγ,ΔGC) = 0.9957). The panel does not encode who can hibernate.
RH8, the v0.4.0 hibernation-bridge expansion and the strongest forward result in the package. A pre-registered null carried from one fuel-switch gene (RH4) to the entire torpor program. 88 promoters were read across deep hibernators (ground squirrels, marmot, hamster, black bear), a torpor-capable primate, daily heterotherms, and non-hibernating controls including a GC-matched rodent. No promoter read tracks hibernation capability; the group differences that exist are the GC differences. The capability is regulatory gating of present genes, not a sequence-level γ threshold.
The panel
Eight genes spanning the furnace (UCP1), the sympathetic command (ADRB3), the fuel switch (PDK4), the mitochondrial coactivator (PPARGC1A), thyroid activation (DIO2), brown-fat identity (CIDEA), the metabolic-state hormone (FGF21) and the insulin-responsive glucose transporter (SLC2A4). Each promoter γ is re-derived offline from the vendored cache, identical scale to every other chapter.
The reads, by gene
For each gene: the hibernator γ range, the non-hibernator γ range, whether the two ranges OVERLAP (they all do), the hibernator-minus-non difference in mean γ and in mean GC, and an exact permutation p (exhaustive over all label splits, no RNG).
| gene | hibernator γ | non-hib γ | ranges overlap | Δγ | ΔGC | exact p |
|---|---|---|---|---|---|---|
| UCP1 | 1.4029–1.4415 | 1.3951–1.4273 | yes | +0.0083 | +0.0077 | 0.282 |
| ADRB3 | 1.3502–1.4525 | 1.4068–1.4976 | yes | -0.0610 | -0.0531 | 0.024 |
| PDK4 | 1.3647–1.4049 | 1.3595–1.4965 | yes | -0.0293 | -0.0238 | 0.169 |
| PPARGC1A | 1.2981–1.4189 | 1.3892–1.4856 | yes | -0.0513 | -0.0414 | 0.097 |
| DIO2 | 1.2814–1.3413 | 1.2617–1.3254 | yes | -0.0061 | -0.0062 | 0.650 |
| CIDEA | 1.3071–1.5123 | 1.4398–1.4987 | yes | -0.0536 | -0.0512 | 0.138 |
| FGF21 | 1.4674–1.5960 | 1.4312–1.6127 | yes | +0.0054 | +0.0058 | 0.880 |
| SLC2A4 | 1.3756–1.5871 | 1.3744–1.5646 | yes | -0.0265 | -0.0233 | 0.554 |
The gap is the GC gap
Across the panel the per-gene γ gaps line up almost exactly with the GC gaps: the cross-gene correlation r(Δγ, ΔGC) is 0.9957, near unit slope. The single nominally low-p gene (ADRB3) fails a Bonferroni correction, is GC-matched, has fully overlapping ranges, and runs biologically backwards (the sympathetic command reads LOWER in hibernators) -- it is the GC confound, not a torpor signal.
Why this is the headline null
This is the present-but-silenced reading (§11) made quantitative and general: the fuel-switch / BAT program is PRESENT across the endotherm set, humans included, and its promoter stiffness does not mark who actually hibernates. Hibernation is REGULATORY gating of a present program, not a property a promoter read can see. A clean negative result is the contribution.
Grades
The reads and the GC confound are [V]: reproducible offline, and the confound is a parameter-free, re-derivable fact (per-gene sign agreement plus the near-unit cross-gene slope). The conclusion that NO promoter marks hibernation is [O]: small n, species are not phylogenetically independent draws, and the regulatory-gating mechanism is cited biology, not derived here.