§13 · feasibility map
The reachability map is the measured spinodal; the yes/no rides on a firewalled drive only (B) scores
A self-contained simulation maps only which switch a drive reaches: reachability is the measured spinodal, and the absolute flip-drive orders by measured promoter γ. It transfers to real autism-gene promoters — each is R19-bistable. Whether a given RNA dose flips a given neuron is a firewalled Δh; only (B), scored against held-out pre/post expression with no tuning, returns yes/no. [V]/[O].
FM1 maps 12 RNA-machinery switches from easiest (AGO2, spinodal 0.6223) to hardest (DGCR8, spinodal 0.7380); a drive at k = 1.05 flips just past the spinodal, k = 0.95 does not. FM2 extends the structure to 10 measured ASD promoters (SCN2A shallowest, PTEN deepest). FM4 shows the self-contained flip is a Δh-replay over all 22 switches, not evidence.
Reachability is the measured spinodal
A drive reaches a switch exactly when it crosses that switch's measured spinodal — there is nothing to assume. Across 12 RNA-machinery promoters the absolute flip-drive is monotone in γ, from AGO2 (easiest, spinodal 0.6223) to DGCR8 (hardest, spinodal 0.7380). A drive scaled to k = 1.05 of the spinodal flips it; the same drive at k = 0.95 does not. The map is geometry, not a prediction that any particular dose lands.
The structure transfers to real autism-gene promoters
The same read applies to 10 measured SFARI ASD-gene promoters: each is an R19 bistable switch, and the flip-drive orders by the measured promoter γ — from the shallowest (SCN2A, γ 1.1982) to the deepest (PTEN, γ 1.5694).
| gene | γ | spinodal | barrier | role |
|---|---|---|---|---|
| SCN2A | 1.1982 | 0.5048 | 0.3589 | voltage-gated sodium channel Nav1.2 (ASD/epilepsy) |
| NRXN1 | 1.3505 | 0.6041 | 0.4560 | neurexin-1 presynaptic adhesion (ASD/schizophrenia) |
| CHD8 | 1.4080 | 0.6431 | 0.4956 | chromodomain helicase DNA-binding 8 (highest-confidence ASD risk) |
| NLGN3 | 1.4366 | 0.6628 | 0.5160 | neuroligin-3 postsynaptic adhesion (X-linked ASD) |
| SYNGAP1 | 1.4463 | 0.6695 | 0.5229 | synaptic Ras-GTPase-activating protein (ASD/ID) |
| FMR1 | 1.4563 | 0.6764 | 0.5302 | fragile-X mental retardation 1 (Fragile-X; translational brake) |
| MECP2 | 1.4951 | 0.7036 | 0.5588 | methyl-CpG-binding protein 2 (Rett; X-linked ASD, chromatin reader) |
| SHANK3 | 1.5207 | 0.7218 | 0.5781 | postsynaptic density scaffold (Phelan-McDermid ASD) |
| TSC2 | 1.5435 | 0.7381 | 0.5956 | tuberous sclerosis 2 / mTOR brake (ASD with TSC) |
| PTEN | 1.5694 | 0.7567 | 0.6158 | PI3K phosphatase (macrocephaly/ASD) |
The honesty gate: a self-contained flip is replay, not evidence
By construction a flip exists at k ≥ 1 for every one of the 22 switches checked, and never at k < 1. That screen is a function of the assumed drive alone — it replays the firewalled Δh rather than testing it, so a self-contained simulation cannot return whether a real cell changed. The only honest crossing is (B): score the predicted flipped-switch set and ordering against held-out measured pre/post expression of real siRNA/saRNA-treated cells, with no tuning of Δh or parameters to the target. Until (B) is run, the per-neuron yes/no stays [O].