# Adversarial review log — enhancer-promoter-looping

**Authoring model:** Sol 6.1. **Adversarial reviewer:** Claude Opus 5.5 (Anthropic). **Date:** 2026-10-14.
**Round:** 2 (2026-10-14).
**Canonical path:** `hsef_v53_final_nine_2026-10-12/enhancer-promoter-looping/enhancer-promoter-looping.review.md`.
**Propagated to:** `hsef_v53_remaining_priority_11_15_2026-10-11/enhancer-promoter-looping/` (verified byte-identical after the copy).

## Scope note — the two files in this folder

This folder holds two markdown documents. `enhancer-promoter-looping.review.md` (the shipped review) is the one
reviewed and corrected here. `current-source-checkpoint.review.md` is a build-tool working artifact holding an
earlier state of the **same** hypothesis, not a separate hypothesis and not a separate review. It was not
adjudicated; it has been given a header note marking it superseded by the shipped file and stating that it carries no
approval. The inventory entry that treated it as its own review is wrong.

**Builder defect to fix upstream:** `current-source-checkpoint.review.md` is a generic working filename. Two
different hypotheses in this delivery (`enhancer-promoter-looping` and `super-enhancer-cell-identity`) ship a file of
exactly that name, which is what invited the misclassification. The builder should emit a slug-specific name, e.g.
`enhancer-promoter-looping.source-checkpoint.md`. Separately, the shipped review links to this file as the "current
source checkpoint" for its credited counts; once the file is renamed, that link needs re-pointing.

## Summary

| Level before | Level after | Confidence before | Confidence after | Critical | Major | Minor | Editorial | Approval |
|---|---|---|---|---|---|---|---|---|
| Β — Associational | Β (unchanged) | Moderate | Moderate (re-derived under lead policy; band stands) | 0 | 1 | 2 | 0 | Approved as revised |

PMIDs checked 60/60, all resolved to the stated first author, title, journal and DOI. Retraction /
expression-of-concern screen: none found. Load-bearing claims spot-checked: 11. Validator
(`validate_review.py` v2.3, HSEF v5.3 tree): **pass** — 0 FAIL, 0 WARN, 31 PASS.

## Findings

| ID | Severity | Location | Original → corrected | Evidence |
|---|---|---|---|---|
| EP-1 | Major | AT A GLANCE (new **Confidence basis** paragraph); `final_score.confidence_rationale` | "density, heterogeneous controls and **incomplete access** set the confidence" → grounded in evidence density and reproducibility, access depth excluded | Lead policy; `hsef-framework.md` § Confidence band. Two-step test applied: access depth was not the only basis — decisive measurements at the edge of their detection limits (kilobase label offsets, finite resolution, 0.5–2-minute reporter delay), one shared tether reagent platform behind the strongest positives, synthetic/ectopic landscapes standing in for native loci, and burst *fraction* standing in for burst *frequency* are all valid determinants — so **Moderate stands**. |
| EP-2 | Minor | Registered claim and scope; REVIEWER NOTES; final-score causal-direction paragraph | "rank43"/"Rank43" → "rank 43"/"Rank 43"; "(step1)"–"(step5)" → "(step 1)"–"(step 5)" | Known renderer defect 4, 7 sites. These glued tokens are cross-references — to the prior popularity rank and to the role-of-mechanism derivation steps — so the defect obscured pointers, not only readability. |
| EP-3 | Minor | Revision checkpoint block | Credited claim count added; historical-vs-credited distinction stated for renderers | The preserved header blockquote reports 60 sources and 65 claims; the score card, ledger and checkpoint all report 57 and 62. The review disclosed the source discrepancy but not the claim discrepancy. The framework makes the provenance header the artifact that travels with the document precisely so a stripped export is recognisable, so a header that over-reports by three sources and three claims is a real export hazard. The prior header is left **verbatim**, as the review's declared revision mode requires; the repair is made in the revision block, which is not part of the preserved set. |

## Scoring logic — checked and upheld

Caps applied per subclaim before thresholds. SC1 is Β under universal cap 2, correct for a spatial-enrichment claim.
The declared `primary-subclaim` rule is applied as written (SC3 is the named gate), and all three rules converge on
Β, which is reported.

The load-bearing move is the one-rung falsification downgrade from Γ to Β on SC3, and it was argued against before
being upheld. It survives because: the exact index directs contrary endogenous predictions to be treated as
falsification; it is applied **once**, not once per contrary paper; the correlated versions of the same
sustained-contact prediction (Sox2 imaging, neural Shh, estrogen loci) are explicitly counted together; and the
review refuses to count cohesin-depletion retained contacts as a second falsification, since retained contacts are
not a contact-ablation test. That refusal is the discipline the framework asks for and the opposite of the
double-counting error. The Α steelman is written and rejected on the record, and the review is careful throughout
that a null within detection limits is not proof that brief contacts never occur.

Momentum "Mechanistic only" is correct and was not inflated: the named 2025/2026 imaging and kinetic reports are
completed publications, which under lead policy can never satisfy "Decisive work underway". Stability "Stable" is
correct by definition where no trigger is named. Divergence "Contested" is permitted without implied-level
arithmetic.

## Rejected findings (checked, no change)

1. **"Best-supported rival" renderer defect.** Rank 1 (topological corral, M2) is `is_nominal: false`; the review's
   own nominal (M1, continuous physical contact necessary) ranks 5. Defect absent.
2. **Four print-vs-e-pub year differences of one** (39721581, 30527662, 39626660, 39626663). Standing convention.
3. **Three title mismatches from PubMed field loss** — the italicised gene names *Sox2* (Alexander) and *Shh*
   (Paliou; Williamson). The review's text is correct. Standing convention.
4. **Bare `PMID31124784` token form** in prose and in the `Study:` ledger field. Systematic house token, not a
   dropped intra-word space.
5. **Preserved header's access figure** ("10 partial-body and 50 complete abstract/status") against the checkpoint's
   "10 partial-body and 47 bounded abstract/status". Same cause as EP-3, now covered by the amended revision block;
   the preserved header is left verbatim.
6. **REVIEWER NOTES vs the Consensus section.** No contradiction: the notes record three summary/original-body aspect
   comparisons and five failed or mismatched retrievals, consistent with the Methods statement and with the
   narrative's correction of one Consensus mis-statement about the Robles-Rebollo design.
7. **Spliced "earlier adversarial approval" sentence.** Absent here — this review is not one of the
   `consensus_consistency_audit_2026-10-09/revised/` copies, and it makes no claim of a prior approval.

## Collective-author screen (both known forms)

Screened for a dropped author leaving a stray leading comma, for a consortium name truncated to its last word so that
it reads as an ordinary surname ("Group et al.", "Network et al.", "Consortium et al."), and for dropped nobiliary
particles. Method: every reference's rendered first author compared against the PubMed record, including records
whose first-author object comes back empty because the article is credited to a consortium.

Result: no collective first author in this list, no truncation, no stray comma; hyphenated and short names intact
("Robles-Rebollo", "Le"). No defect.

## Residual limitations

Access depth unchanged: ten partial-body examinations, the remainder abstract/status, no complete-paper or supplement
audit. The Du and Narita correction bodies remain unexamined, as the review's own knowledge-gap list states.
Partial-body figures could not be re-verified beyond abstracts and are not contradicted by any record examined.

## Line-ending note

The edit pass initially normalised these files from CRLF to LF. That was unintended and has been reversed: the review,
score card and checkpoint are back on CRLF, matching their pre-edit state and the rest of the package, and the hashes
below are the final CRLF files.

## Changed files (identical in both folders)

| File | sha256 | bytes |
|---|---|---|
| enhancer-promoter-looping.review.md | fdbda6b5303462c28aab8bc3f9afa73781e8a2091fa0091ac6fed06798156c8a | 97658 |
| enhancer-promoter-looping.scorecard.json | 0dd4d79da86d18e1ca6353f1efe9eea49f3692f6dba1b45cdce77d9093493b7a | 278691 |
| validation-report.json | 5347ff6e45b57498a74b6f581cb204416dc63e586ba1d41bb05536ba73d9d545 | 7084 |
| SHA256.txt | 81bb774a8cc70e3c7461024cd0a2c84c25d133ffe6bd529718e3e168df43b319 | 213 |
| current-source-checkpoint.review.md (header note only) | 733cd426087483cf3116973271e0efa64b6f36732aa0ebd4cbb346f532b52c81 | 86697 |

Propagation verified: all 14 files shared between
`hsef_v53_final_nine_2026-10-12/enhancer-promoter-looping/` and
`hsef_v53_remaining_priority_11_15_2026-10-11/enhancer-promoter-looping/` hash-match after the copy, as they did
before the edits. Pre-edit copies are in `sol61-original/` in **both** folders. The two divergent copies
(`hsef_v53_popularity_ranks_46_50_2026-10-09/` and the `prior-edition/` subfolder) were not touched.

**Approval:** Following adversarial review and the corrections listed above, the revised document is approved. — Claude Opus 5.5 (Anthropic), 2026-10-14.