FINAL AUDIT REPORT
Protein interaction-response release v1.0.4
Layout audit: 16 September 2026
Manuscript: From Microscopic Ensembles to Validation-Safe Transfer-Energy Compression

SCOPE
This revision corrects presentation and document structure. Scientific inputs,
numerical outputs, claim registries and analysis programs are preserved from
the complete scientific replay of 15 September 2026. The original audit and
full execution transcript remain in verification/. Full stochastic campaigns
were not rerun solely for this layout revision.

LAYOUT
- Six numbered main sections, with consistent subordinate headings.
- Main pages 1--18; supporting pages S1--S22. No line-number clutter.
- Sixteen numbered displayed equations in each document: (1)--(16) and
  (S1)--(S16); three main and eighteen supporting tables.
- All four publication figures redrawn; duplicate captions removed.
- Eight diagnostic figures regenerated with legible labels and vector exports.
- Vector PDF/SVG and high-resolution PNG supplied for all twelve figures.
- Repeated table headers, intact table rows in DOCX, consistent column widths.
- Two malformed source-table rows repaired in SI S2.1; endpoint definitions
  and units agree with the source registry and main Table 1.
- All 18 main and 22 SI PDF pages visually inspected, as were all 18
  manuscript and 22 SI DOCX-render pages. All twelve figures inspected.
- PDF fonts embedded; no missing-glyph or overfull-box warnings; all equation
  tags and all 46 claim identifiers present. No tracked changes or comments.
- PDFs are independently typeset from supplied LaTeX sources. DOCX files are
  editable versions of the same content, with independent pagination.
- The unchanged one-page cover letter is retained. This is a clean reading
  layout; the earlier audit's line-number/double-spacing description applies
  only to the preceding edition.

SCIENTIFIC PRESERVATION
121 pre-existing reproducibility files are byte-identical. The only changed
pre-existing files inside reproducibility/ are eight figure PNGs and one
plotting script. Added files provide vector assets and a plotting producer.
SHA-256 evidence is in verification/SCIENTIFIC_CONTENT_PRESERVATION.json.
The 33 references and targeted recent-literature comparison are retained.

INTEGRITY
Both manifests are rebuilt after synchronization. verify_release.py checks
release integrity, predictive contracts, interaction geometry and independent
exact representation certificates. Its final execution record is supplied,
with an additional fresh-extraction check accompanying the distributed ZIPs.
The scientific package nested inside the submission package is byte-identical
to the separately supplied scientific ZIP. External SHA-256 files identify
both archives.

UNCHANGED INFERENCE BOUNDARIES
The historical D≈3 coefficient pattern is retrospective and basis-specific.
Its five-million-permutation plus-one p = 1/5,000,001 is conditional on the
selected historical coding; it does not adjust for descriptor/constant
selection and is not a test of intrinsic physical dimension. Changing only
Cys coding from one to zero changes template distance from 0.017693128924 to
0.0774229837224. The p-value is not transferred to that changed design.
Moon-Fleming retains CONDITIONAL status and fixed-class stress p = 0.1127177456.
Hessa and Moon row-level uncertainty gaps remain open. Shared prediction
does not identify a unique physical generator. Prospective outcomes remain
unobserved. No numerical outcomes or new experimental data were invented.
No journal submission or external public deposition was performed.
