# Change-log — EoE pMHC-II manuscript repositioning

**Source:** `EoE_Final_Manuscript_Main.docx` (v9de16479, 9 Jul 2026), title *"Rational Design of HLA-Restricted pMHC-II Nanoparticle Therapeutics for Eosinophilic Esophagitis."*
**Result:** `EoE_manuscript_personalized.docx`, retitled *"A patient-specific antigen-selection engine for personalized pMHC-II tolerance therapy in eosinophilic esophagitis."*
**Driver:** align the manuscript with the updated `scientific_plan.md` and `business_plan.md` (both 10 Jul 2026, Tolera Bio), which had moved ahead of the paper; and reconcile internal factual inconsistencies. See `manuscript_reconciliation.csv` for the item-by-item authority.

## What changed and why

1. **Thesis: fixed-HLA design → personalized antigen-selection engine.** The paper was built around three fixed epitopes on a single allele (DRB1\*07:01), with multi-HLA expansion listed as future work. The plans make *per-patient, HLA-driven antigen selection* the central advance (the PACT™ shared-backbone/swappable-cassette concept). The manuscript now leads with the engine; the specific validated constructs are presented as a worked example. *Why: the plans' core intellectual contribution was absent from the paper.*

2. **New Figure 1 built from real prediction data.** A multi-HLA presentation-hierarchy figure was computed fresh (`pmhc_ii_candidates_msw.csv`): β-casein, α/β-gliadin, β-conglycinin × 5 HLA-DRB1 alleles = **924 15-mers → 4,640 predictions → 832 strong binders**, with per-allele load spanning **13–352**. This is the quantitative basis for per-patient selection. *Why: the repositioned thesis needed a data figure; the old Fig 1 (single-allele epitope ranking) did not carry it.*

3. **Dairy antigen corrected: β-lactoglobulin → β-casein.** The lead epitope core FAQTQSLVY is β-casein (UniProt P02666, aa 67–75, inside the aa 59–78 window); it is **absent from β-lactoglobulin**. The manuscript and the earlier bulk-scan file had mislabeled it. *Sequence-verified. User confirmed β-casein canonical.* Milk was re-scanned on β-casein so Figure 1 matches the lead.

4. **Nanoparticle surface coverage corrected: 5.97% → 1.4%.** The spec table (`table3_nanoparticle_specs.csv`) gives 17.5 nm² of a ~1,257 nm² 20 nm-sphere surface = 1.4%; the manuscript's 5.97% was not reproducible. *Arithmetic-verified.*

5. **HLA "discrepancy" retired (not an error).** An earlier note flagged the candidate data as computed only on DRB1\*01:01 vs. the DRB1\*07:01 claim; the full data actually spans 5 alleles **including \*07:01**. Reframed as available multi-HLA depth, not a fix.

6. **Egg dropped; wheat/soy substituted (user decision).** The prior bulk file included egg (ovomucoid, Gal d 1) the author had not reviewed. Per user direction, Figure 1 was rebuilt on milk/wheat/soy — matching the existing dairy/wheat/soy structural panel. Egg is noted in Limitations as unfolded/unvalidated.

7. **Biomarker-first framing added.** Discussion and a new development-cascade narrative are organized around the PD-biomarker spine (structural gate → biochemical → tetramer → IL-10/Tr1 signature → in-vivo → human PD) and the five ASIT-landscape lessons (Nexvax2, teplizumab/Tzield, TAK-101/KAN-101, GAD-alum) the plans are built on. *Why: absent from the paper; the field's decisive lesson.*

8. **Hackathon budget removed.** The manuscript's "$275–457k, 5-phase, 18–24 mo" roadmap (~20× below the business plan's $8M Seed → $35M Series A, and inappropriate in a research manuscript) was replaced with a qualitative, biomarker-gated development cascade and companion-diagnostic co-development.

9. **Construct/panel framing aligned.** Structural results stated as the **9-construct panel, ipTM 0.87–0.90, 15/15 in-groove** framing used by the plans, with the specific complexes as worked examples.

10. **Honesty additions.** Explicit Limitations: predicted binders ≠ immunodominance/processing/tolerance; wheat/soy/egg as computational priors; single worked allele for structural validation; β-casein tetramer/TCR (eoeTCR-4) attribution flagged for independent citation verification (matching the plans' own caveat); pMHC modality clinically unproven.

## Reused unchanged
- **Figure 2** (structural validation, `figure2_structural_validation.png`) and **Figure 3** (nanoparticle architecture, `figure3_nanoparticle_architecture.png`) — reused with corrected captions (surface coverage).

## Supplemental (rebuilt to match the repositioned main text)

`EoE_Final_Manuscript_Supplemental.docx` (v1adae273) predated the repositioning and carried the same errors as the old main text. Rebuilt as `EoE_manuscript_personalized_supplemental.docx`:
- **Table S1** — recast around the antigen-selection engine: β-casein (not β-lactoglobulin) dairy lead with the correct FAQTQSLVY-spanning peptide (QDKIHPFAQTQSLVY, aa 61–75, 131.9 nM) and the distinct top β-casein core (aa 174–188, 13.4 nM); wheat/soy leads from the fresh mhcnuggets scan; DRB1\*07:01 shown as one of five scanned alleles; eoeTCR-4 tetramer claim now flagged pending citation verification. All predictions now cite one predictor (mhcnuggets), consistent with Figure 1.
- **Table S2** — structural metrics retained; framed against the ipTM ≥ 0.87 acceptance gate.
- **Table S3** — surface occupancy corrected 5.97% → **1.4%**; PACT™ shared-backbone manufacturing row added.
- **Table S4** — the $275–457k 5-phase budget roadmap **replaced** with the biomarker-gated development cascade (no fundraising figures in a research supplement).
- **Table S5** — risk matrix realigned to the plan's scientific risks (antigen heterogeneity, reactogenicity, no validated PD biomarker, N-of-1 CMC).

## Submission-readiness pass (citations + references)

**eoeTCR-4 citation — WITHDRAWN.** Exhaustive PubMed + OpenAlex search found no publication for an "eoeTCR-4" clone, the FAQTQSLVY motif, a tetramer-validated β-casein/HLA-DRB1\*07:01 EoE epitope, or any β-casein CD4 epitope map matching the claim. The claim was the manuscript's only asserted *experimental* (non-computational) anchor. It has been **downgraded to a computational prediction** throughout (Abstract disclaimer, Results, Limitations, Table S1), with the withdrawal stated explicitly. The milk lead is now motivated by real, verified literature on milk-responsive CD4⁺ T cells in EoE (refs 5, 6), not by a validated epitope–TCR triad. Tetramer/TCR validation is named as the Stage-2B wet-lab milestone.

**Reference list — 16 references, all DOI-verified.** Every DOI was checked against CrossRef/OpenAlex to confirm it resolves to the cited paper (one initial Nexvax2 DOI guess resolved to the wrong trial — AMG 714 — and was replaced with the correct Goel 2017 EBioMedicine paper). Numbered [1]–[16], threaded through the main text and supplemental (shared numbering). Full list in `manuscript_references.csv`. Real primaries now cited: Dellon AGREE 2018, dupilumab NEJM 2022, Morgan Sci Immunol 2021, EoE food-specific TCR Allergy 2023, milk-T2 EoE 2017/2024, teplizumab NEJM 2019, Nexvax2 EBioMedicine 2017, TAK-101 Gastroenterology 2021, Navacim Nature 2016, neoantigen vaccine Nature 2023, AlphaFold + AlphaFold-Multimer, MHCnuggets, IEDB 3.0.

## Remaining before submission (journal-dependent)

- **Author list / correspondence / affiliations** — still placeholder-level; needs finalizing.
- **Journal formatting** — reference style (currently a plain verified list), figure-callout conventions, word/section limits depend on target venue. The prior submission index named *Nature Immunology / Nature Communications*; confirm target before a formatting pass.
- **Scope confirmation** — this is a computational design/framework paper (all wet-lab data is predicted, disclosed as such). Confirm the target venue accepts in-silico methodology papers.

## Deliverables
- `EoE_manuscript_personalized.docx` — repositioned manuscript, 3 embedded figures (300 dpi).
- `EoE_manuscript_personalized_supplemental.docx` — rebuilt supplemental (Tables S1–S5).
- `EoE_manuscript_personalized.md` — markdown source.
- `fig_personalization_engine.png` — new Figure 1.
- `pmhc_ii_candidates_msw.csv` — fresh milk/wheat/soy × 5-allele scan (4,640 rows).
- `manuscript_reconciliation.csv` — every factual fix with evidence and source (now incl. eoeTCR-4 withdrawal + references).
- `manuscript_references.csv` — 16 DOI-verified references with resolution check.
- `manuscript_changelog.md` — this memo.