PROTEIN COMPRESSION AUDIT - REPRODUCIBILITY REPORT

1. EXACT STRUCTURAL CERTIFICATION
structural/baseline replay: 24/24 PASS
reference: 1023 syntax -> 415 semantic -> 262 low-capacity -> 131 protocol-safe
irredundant control: 255 syntax -> 255 semantic
observer-safe D8: 103 semantic -> 93 low -> 47 safe
Moon pH3.8: 207 semantic -> 161 low -> 89 safe
Exact safe-state registry: results/EXACT_SAFE_CLASS_REGISTRY.json

2. CORE END-TO-END REPLAY
core replay: 12/12 PASS
  H1 WW_Octanol: 0.8245398818153601
  H1 WW_Interface: 0.5294248860882051
  H1 Hessa: 0.5860847514007514
  H1 Moon_Fleming: 0.21322753325070543
  H1 Eisenberg: 0.6754112488488581
  H1 Fauchere_Pliska: 0.7747511135018753
  H2a cluster_weighted4: 0.5490920693730572
  H2a dependency_reduced3: 0.5756109186180025
  chemotype WW_Octanol: -1.9766598014269015
  chemotype WW_Interface: -1.3564044007128349
  chemotype Hessa: 0.27673487019151877
  chemotype Moon_Fleming: 0.288529290317879

3. PRIMARY PIPELINE-EXACT H1 PERMUTATION NULL (50,000 PER ENDPOINT)
  WW_Octanol: exceed=0/50000; p+=1.9999600008e-05; Holm=7.9998400032e-05; q95=0.086233; q99=0.325838; observed algorithm identity=True
  WW_Interface: exceed=28/50000; p+=0.000579988400232; Holm=0.00115997680046; q95=0.058071; q99=0.280523; observed algorithm identity=True
  Hessa: exceed=10/50000; p+=0.000219995600088; Holm=0.000659986800264; q95=0.106834; q99=0.297037; observed algorithm identity=True
  Moon_Fleming: exceed=1143/50000; p+=0.0228795424092; Holm=0.0228795424092; q95=0.102845; q99=0.314373; observed algorithm identity=True

3a. ENDPOINT-ISOLATED PIPELINE-EXACT FULL REPLAYS
endpoint-isolated full replay: 4/4 exact ledger match; observed algorithm identity PASS for all four endpoints.

4. AUXILIARY BEST-FIXED-CLASS STRESS NULL (50,000 PER ENDPOINT)
  WW_Octanol: exceed=0/50000; p+=1.9999600008e-05; q95=0.307957; q99=0.474150
  WW_Interface: exceed=132/50000; p+=0.00265994680106; q95=0.281426; q99=0.434980
  Hessa: exceed=35/50000; p+=0.000719985600288; q95=0.267225; q99=0.408706
  Moon_Fleming: exceed=5635/50000; p+=0.112717745645; q95=0.306805; q99=0.446977
No analytical dominance relation between the primary adaptive statistic and this fixed-class stress statistic is assumed.

5. CHARGE-MAGNITUDE SENSITIVITY
charge-sensitivity replay: 5/5 PASS
WW Octanol Generic11+|FormalCharge| Q2=0.816631140208802; exact random-sign p=0.8852386474609375.

6. TYPED CLAIM BOUNDARIES
  H1_WW_OCTANOL=PIPELINE_NULL_CONTROLLED_EMPIRICAL
  H1_WW_INTERFACE=PIPELINE_NULL_CONTROLLED_DEPENDENT_EMPIRICAL
  H1_HESSA=PIPELINE_NULL_CONTROLLED_EMPIRICAL
  H1_MOON_FLEMING=QUALIFIED_PIPELINE_NULL_CONTROLLED
  H2B_UNIQUE_GEOMETRY=NONIDENTIFIABLE_AT_CURRENT_DESIGN_RESOLUTION
  UNIVERSAL_PREDICTIVE_SUPERIORITY=REJECTED
  CHEMOTYPE_GENERALIZATION=NO_GO_FOR_BROAD_EXTRAPOLATION

7. SCOPE
Exact structural certification is authoritative for quotienting and validation-domain rank safety. The numerical core consumes that exact registry. The verifier checks package integrity and claim consistency; it is not an independent experimental replication of every stochastic ledger.

8. FULL DETERMINISTIC STACK
reproduce_all.py: PASS (24/24 structural+matched, 12/12 core, 5/5 charge sensitivity, 13/13 mutation guards rejected).

9. CURRENT COMPLETE STOCHASTIC REPLAY
All H1 adaptive and fixed-class stress campaigns were freshly executed.
H2 alignment: two 10,000-draw campaigns; both zero exceedances, with current
quantiles verified. H2 local sensitivity: 9,000 explicitly specified draws.
WW uncertainty: two 10,000-perturbation campaigns, both matched.
See the root FINAL_AUDIT_REPORT.txt for precise H2 recomputation provenance.
Current Figures 5--6 use the current H2 ledgers, not superseded summaries.
