# Dry-run — monogenic / genetic archetype (SMA, HD, splice)

Validation that the **archetype-adaptive** build works on genetic disease — the class the
original EoE-derived pipeline broke on. Planning logic only (no compute); confirms the pipeline
branches correctly and routes to genetics-appropriate modalities.

## Case A — Spinal Muscular Atrophy (SMA): monogenic, recessive-null LoF, sparse data

| Step | Result |
|---|---|
| Stage 1b archetype | **monogenic** (ClinGen definitive, 1 causal gene = SMN1) |
| Data density | **sparse** (0 omics cohorts, 0 registered trials in the triage inputs) |
| Stage 3 path | `natural_history_plus_analogous_precedent` (not a competitor matrix) |
| Stage 4 path | `genetics_led_confirmation` (**not** omics DE meta-analysis) |
| Stage 5 modality | **gene_replacement** (alternatives: enzyme_replacement, readthrough, base/prime editing) |
| Delivery vehicle | AAV serotype/tropism, ~4.7 kb payload limit, or LNP |

The pipeline correctly recognizes that SMN1 is *already the target*, so Stage 4 confirms
mechanism rather than discovering a target, and Stage 3 pivots to natural history + analogous
precedent because there is no competitive pipeline to map in the triage scenario.

## Case B — Huntington disease (HD): monogenic, repeat-expansion (toxic gain)
- Archetype **monogenic**, data density **moderate**.
- Stage 5 modality → **aso_knockdown** (alternatives: siRNA knockdown, CRISPR nuclease).

## Case C — splice-variant monogenic
- Stage 5 modality → **splice_switching_aso**.

## Landscape cross-check (the router reproduces reality)
The routed modalities match the **actual approved / clinical-stage** therapies for these
diseases — evidence the mechanism→modality tree is grounded, not arbitrary:
- SMA → onasemnogene abeparvovec (**gene replacement**, AAV9) and nusinersen (**splice-switching
  ASO**). Both fall out of the SMA routing (gene_replacement lead; splice case → SSO).
- HD → tominersen (**ASO knockdown**), matching Case B.

## Result
- Archetype classifies correctly; data triage selects the genetics-led path for sparse monogenic
  disease instead of the omics-discovery path.
- The mechanism→modality router returns genetics-appropriate modalities across LoF-recessive,
  repeat-expansion, and splice mechanisms — **none defaulting to the EoE modality set**
  (vaccine / antibody / ligand-trap).
- ARCHETYPE LOCK and the Stage-3/4 branches fire as designed.

**Conclusion:** the pipeline now covers the rare/monogenic majority. The disease archetype,
mechanism, and modality are all data-resolved parameters gated by human confirmation — "any
disease" is one adaptive pipeline, not a common-disease tool with a genetic disclaimer.

> Caveat: mechanism of pathogenesis here was supplied as a confirmed input. In a live run it is a
> clinical claim that must be human-validated at ARCHETYPE LOCK (grounded in ClinVar/ClinGen/
> gnomAD/literature) before it drives modality choice.