§18 · application: lever map
Every modality bins by channel: a SET edit or a fixed-γ drive — and a finite γ* mandates the edit
SET edits (knockout/base/prime) move the spinodal and barrier and are not drive-reversible; CRISPRa leaves γ untouched and reverts, so the substrate re-classifies it as Lever B. The Lever-B sub-types (siRNA −, ASO-splice a fixed-γ coordinate change, saRNA +, miRNA-sponge +) all act at fixed γ and revert. The decision boundary hpath*(γ) = hcap − spinodal(γ) falls and crosses zero at a finite γ* (1.4598); beyond it even a zero-hold switch needs Lever A. [V].
On DICER1 (γ = 1.3981, spinodal 0.6363, barrier 0.4887) a knockout drives γ to 0.0699 (threshold gone) while CRISPRa leaves γ fixed and reverts. On TARBP2 (γ = 1.5122) every Lever-B sub-type leaves the spinodal and barrier byte-identical and reverts on withdrawal. With a declared cap h_cap = 0.6789 the boundary crosses zero at γ* = 1.4598 (analytic) / 1.4605 (swept), and re-dosing above w* = 0.0553 holds the corrected state.
Lever A sub-types move the threshold; CRISPRa is really Lever B
On DICER1 (γ = 1.3981, spinodal 0.6363, barrier 0.4887), knockout, base-edit and prime-edit each change γ itself — the threshold moves and no drive restores it (true SET edits). CRISPRa, despite the “activation” label, leaves γ (and so the spinodal and barrier) untouched and reverses on withdrawal, so the substrate re-classifies it as a drive — Lever B.
| modality | channel | γ after | threshold moved | reversible by drive |
|---|---|---|---|---|
| knockout | SET (gamma) | 0.0699 | yes | no |
| base_edit | SET (gamma) | 1.3581 | yes | no |
| prime_edit | SET (gamma) | 1.2981 | yes | no |
| CRISPRa | DRIVE (h) at fixed gamma | 1.3981 | no | yes |
Lever B sub-types: one channel, different signs
On TARBP2 (γ = 1.5122, spinodal 0.7158, barrier 0.5717) every Lever-B sub-type leaves the spinodal and barrier byte-identical and reverts on withdrawal; only the sign differs. siRNA drives −, saRNA +, a miRNA-sponge + (net de-repression), and an ASO splice-switch is a fixed-γ A4-coordinate change rather than an expression drive.
| modality | axis | sign | reversible | γ untouched |
|---|---|---|---|---|
| siRNA | expression (drive h) | - | yes | yes |
| ASO_splice | structure (A4 coordinate), gamma fixed | coordinate | yes | yes |
| saRNA | expression (drive h) | + | yes | yes |
| miRNA_sponge | expression (drive h) | + (net, de-repression) | yes | yes |
The lever choice is an explicit boundary curve
Declare a tolerable cap h_cap = 0.6789 (the spinodal at the anchor γ = 1.4598, not tuned to an outcome). The drive a cap-bounded Lever B must clear is hpath*(γ) = h_cap − spinodal(γ); it falls with γ and crosses zero at a finite γ* = 1.4598 (analytic), 1.4605 (swept). Beyond γ* even a zero-hold switch cannot be cleared by a tolerable drive and demands Lever A.
| γ | h_path*(γ) |
|---|---|
| 1.3276 | 0.0901 |
| 1.3808 | 0.0544 |
| 1.4339 | 0.0180 |
| 1.4871 | -0.0191 |
| 1.5402 | -0.0569 |
| 1.5934 | -0.0953 |
The boundary curve is read; the cap's absolute value is [O] (it reuses the O-7 tolerability logic).
Durable correction without an edit
Re-dosing a Lever-B drive each cycle — with the per-cycle loss (0.5527) measured from the switch's own relaxation on TARBP2 — holds the corrected state above a critical re-write rate w* = 0.0553: at 1.5×w* the steady state sits at 0.1500 (maintained), at 0.5×w* it falls to 0.0500 (fades), and a single dose with no re-write decays to 0.0000. So a reversible drive can correct durably with γ never touched — the edit-free durability that closes the lever map. The persistence condition is read; the absolute re-dosing schedule is [O].