Scientific Reports revision: geometry, matched graph context, and molecular support

.gitattributes

@@ -1,3 +1,7 @@
 # Hash-covered scientific evidence must retain its supplied bytes on every platform.
 checkpoints/legacy/** -text
 manuscript/revised/** -text
+
+# Preserve new journal evidence hashes across checkout platforms.
+experiments/journal_revision/** -text
+manuscript/journal_revision/** -text

.github/workflows/publication.yml

@@ -4,6 +4,14 @@ on:
     paths:
       - 'manuscript/**'
       - 'src/graphpathology/*publication*'
+      - 'src/graphpathology/extension*.py'
+      - 'src/graphpathology/editorial.py'
+      - 'src/graphpathology/backbone*.py'
+      - 'configs/journal_revision/backbone_graph_comparison/**'
+      - 'experiments/journal_revision/backbone_graph_comparison/**'
+      - 'experiments/editorial/**'
+      - 'experiments/journal_revision/transformer_egnn/**'
+      - 'configs/journal_revision/transformer_egnn/**'
       - 'docs/REPRODUCTION_TARGETS.csv'
       - 'configs/validation-environment.txt'
       - '.github/workflows/publication.yml'
@@ -26,9 +34,19 @@ jobs:
         run: sudo apt-get update && sudo apt-get install -y texlive-latex-extra texlive-fonts-recommended lmodern poppler-utils
       - run: python -m pip install -r configs/validation-environment.txt
       - run: python -m pip install -e '.[test]'
+      - name: Verify committed extension evidence before document builds
+        run: python -m graphpathology.extension verify --committed-only
+      - name: Verify editorial scientific-source preservation
+        run: python -m graphpathology.editorial check --committed-only --output outputs/editorial-integrity.json
+      - name: Verify matched-backbone candidates, selection and paired evidence
+        run: python -m graphpathology.backbone_selected_verify --committed-only --output outputs/backbone-comparison-integrity.json
       - run: python -m graphpathology publication --output outputs/publication
       - run: python -m graphpathology build-documents --publication outputs/publication --output outputs/paper
       - run: python -m graphpathology arxiv --output outputs/arxiv
+      - name: Build separate journal revision from committed exports (no training or research downloads)
+        run: python -m graphpathology.journal paper --build-output outputs/journal
+      - name: Build current Scientific Reports manuscript from maintained sources
+        run: python -m graphpathology.extension paper --build-output outputs/journal-extension
       - uses: actions/upload-artifact@v4
         if: always()
         with:
@@ -40,3 +58,10 @@ jobs:
             outputs/paper/*.pdf
             outputs/arxiv/ARXIV_REPORT.json
             outputs/arxiv/*.zip
+            outputs/journal/*.pdf
+            outputs/journal-extension/*.pdf
+            outputs/journal-extension/EXTENSION_BUILD_REPORT.json
+            outputs/extension-validation-committed.json
+            outputs/editorial-integrity.json
+            outputs/backbone-comparison-integrity.json
+            manuscript/journal_revision/JOURNAL_BUILD_REPORT.json

.gitignore

@@ -7,3 +7,29 @@ outputs/
 work/
 checkpoints/models/*
 !checkpoints/models/README.md
+
+# New journal-v2 trained binaries/caches are separate from released evidence.
+checkpoints/journal_revision/
+experiments/journal_revision/results/**/.evaluation.lock
+
+# Full extension per-gene exports are preserved in a hashed, release-ready author bundle.
+# Keep panels, support bins, completion records and manifests reviewable in Git.
+/experiments/journal_revision/transformer_egnn/evaluation/**/GENES.csv.gz
+/experiments/journal_revision/transformer_egnn/evaluation/**/MOLECULAR_GENES.csv.gz
+/experiments/journal_revision/transformer_egnn/bassiouni_anchors/**/GENE_METRICS.csv.gz
+
+# TeX intermediates in the new journal version; keep final named PDFs and .bbl sources.
+/manuscript/journal_revision/transformer_egnn/main.pdf
+/manuscript/journal_revision/transformer_egnn/article_only.pdf
+/manuscript/journal_revision/transformer_egnn/supplementary.pdf
+/manuscript/journal_revision/transformer_egnn/*.aux
+/manuscript/journal_revision/transformer_egnn/*.log
+/manuscript/journal_revision/transformer_egnn/*.out
+/manuscript/journal_revision/transformer_egnn/*.fls
+/manuscript/journal_revision/transformer_egnn/*.blg
+/manuscript/journal_revision/transformer_egnn/*.pass*.txt
+
+# New matched-backbone gene exports are retained in the author-review bundle.
+/experiments/journal_revision/backbone_graph_comparison/evaluation/**/GENES.csv.gz
+
+/experiments/journal_revision/backbone_graph_comparison/primary_evaluation/**/GENES.csv.gz

README.md

@@ -3,18 +3,22 @@
 **GraphPathology enables geometry-aware reconstruction of spatial transcriptomics from sparse molecular measurements**
 Truong-Son Hy, University of Alabama at Birmingham.
 
-[Combined paper + SI](manuscript/GraphPathology_Combined_Manuscript.pdf) · [Article only](manuscript/GraphPathology_Article_Only.pdf) · [Supplement](manuscript/GraphPathology_Supplementary.pdf)
+[arXiv v1 combined paper + SI](manuscript/GraphPathology_Combined_Manuscript.pdf) · [Article only](manuscript/GraphPathology_Article_Only.pdf) · [Supplement](manuscript/GraphPathology_Supplementary.pdf)
 
-GraphPathology separates contextual image representations, reconstruction from observed same-section molecular anchors, and image–molecular fusion. Graph neural predictors are development comparisons; the external headline results come from molecular interpolation. The tested fixed image contribution did not reproducibly improve external interpolation. V1/V2 differ in weighting and scoring support as well as geometry. External images are downsampled. A potential fivefold reduction concerns selective acquisition and proportional variable location costs, not proven total assay savings or learned long-range correlations.
+GraphPathology separates contextual image representations, reconstruction from observed same-section molecular anchors, and image–molecular fusion. Image and graph predictors are evaluated separately from molecular reconstruction; the original external headline results come from molecular interpolation. The tested fixed image contribution did not reproducibly improve external interpolation. V1/V2 differ in weighting and scoring support as well as geometry. External images are downsampled. A potential fivefold reduction concerns selective acquisition and proportional variable location costs, not proven total assay savings or learned long-range correlations.
 
 | Cohort | Role | Panel |
 |---|---|---|
 | HEST NCBI783 / NCBI785 | Fitting | 50 ordered genes |
 | HEST TENX95 / TENX99 | Validation / primary development evaluation | 50 |
 | HER2ST A1–H1 | Transportability, then calibration development on the same patients | 40-gene intersection |
 | Wu 1142243F, CID4465, CID4535 | Calibration | 50 |
-| Wu 1160920F, CID4290, CID44971 | Fresh evaluation | 50 |
-| Bassiouni, first section for each of 22 patients | Fresh replication, no refitting | 50 |
+| Wu 1160920F, CID4290, CID44971 | Held-out at v1; now inspected | 50 |
+| Bassiouni, first section for each of 22 patients | Held-out at v1, no refitting; now inspected | 50 |
+
+## Post-v1 Scientific Reports journal revision
+
+[Journal experiments and reproduction](docs/JOURNAL_REVISION.md) · [Separate journal sources/PDFs](manuscript/journal_revision/README.md) · [Versioned exploratory protocol](experiments/journal_revision/PROTOCOL_V1.md). New real-data matched-information and identical-target analyses, patient uncertainty and controlled neural training are separate from the arXiv v1 package. The later exact GEO range recovery completed the 22-patient Bassiouni matched spatial comparisons; the earlier source-transfer failure is retained in the historical reports. See the journal status for precise completed/blocked/skipped work. Previously inspected cohorts are not a new blind validation.
 
 ## Install and CPU quick start
 
@@ -52,7 +56,7 @@ python -m graphpathology doctor --require-models
 python -m graphpathology infer --config configs/replay.example.json --dry-run
 ```
 
-The original fitted states are still missing, including all three selected primary heads. These commands report the required paths and checksums and never fall back to new/random weights. The maintained inference route starts from aligned embeddings and normalized targets; it is not a complete raw-cohort research orchestrator. Full raw-data ingestion, extraction, training and evaluation remain separately scoped future work, with the dependency order and prerequisites documented in [REPRODUCING.md](docs/REPRODUCING.md) and [DATA_AND_MODELS.md](docs/DATA_AND_MODELS.md). New training cannot substitute for replay of the original fits.
+The original fitted states are still missing, including all three selected primary heads. These commands report the required paths and checksums and never fall back to new/random weights. The maintained inference route starts from aligned embeddings and normalized targets; it is not a complete raw-cohort research orchestrator. The separate explicit [journal research workflow](docs/JOURNAL_REVISION.md) now supports public-data acquisition, frozen feature extraction, controlled new training and evaluation. New journal fits have independent identities and cannot substitute for replay of original fits. The original bounded inference route retains its existing contract.
 
 ## Organization and provenance
 
@@ -72,3 +76,11 @@ This checks archived bytes and declared registry prefixes without running histor
 ## Citation and terms
 
 Use the manuscript title and author above; [CITATION.cff](CITATION.cff) supplies citation metadata. No journal acceptance or DOI is asserted. HEST, HER2ST, Wu, Bassiouni and upstream model weights retain their own terms. The original repository had no root software LICENSE; this revision does not invent a license grant or redistribute raw tissue data or upstream weights.
+
+### Post-v1 bounded Transformer / EGNN extension
+
+The separate [extension documentation](docs/GRAPH_TRANSFORMER_EGNN.md) and [scientific report](experiments/journal_revision/transformer_egnn/SCIENTIFIC_REPORT.md) describe 48 new controlled fits and 78-model evaluation in 26 previously inspected patients. Exact GEO range recovery completed the 22-patient matched spatial baseline study. The primary neural models did not improve same-information IDW; all negative findings, controls and patient effects are retained. The versioned journal manuscript is under [manuscript/journal_revision/transformer_egnn](manuscript/journal_revision/transformer_egnn/). Earlier arXiv, journal and A–E evidence remains unchanged. Full gene exports/checkpoints have hashed author-local bundles and a release-ready manifest; raw datasets/model binaries are excluded from Git. Draft PR #3 requires author review before merge.
+
+### Current matched-backbone Scientific Reports revision
+
+The maintained [journal manuscript](manuscript/journal_revision/transformer_egnn/) centers the controlled 22-patient measurement-placement result, with an expanded Introduction and a separate Related Work section. The [matched-backbone follow-up](docs/MATCHED_BACKBONE_COMPARISON.md) separates graph-versus-local prediction within ResNet18 and Phikon from the value beyond identical molecular support. All previous empirical findings remain; the conceptual information-boundary schematic is archived outside the active paper. New fits and validation-guided selection are separate from original states and earlier optimization regimes. See the [exploratory protocol](experiments/journal_revision/backbone_graph_comparison/PROTOCOL.md) and study manifests for actual execution and availability. Author-local weights and numerical bundles are not public deposits. The author approved the final AI disclosure and authorized merging [PR #3](https://github.com/HySonLab/GraphPathology/pull/3); the PR records its current status.

configs/journal_revision/backbone_graph_comparison/protocol.json

@@ -0,0 +1,135 @@
+{
+  "created_utc": "2026-11-01T16:24:58.305126+00:00",
+  "classification": "exploratory_follow_up",
+  "base_commit": "ab0811134bb1d62719206ab11a43bd7c4dd5d813",
+  "backbones": [
+    "resnet18",
+    "phikon"
+  ],
+  "configurations": [
+    {
+      "model": "local",
+      "control": "local"
+    },
+    {
+      "model": "MEAN",
+      "control": "spatial"
+    },
+    {
+      "model": "GCN",
+      "control": "spatial"
+    },
+    {
+      "model": "SAGE",
+      "control": "spatial"
+    },
+    {
+      "model": "GCN",
+      "control": "shuffled"
+    },
+    {
+      "model": "SAGE",
+      "control": "shuffled"
+    }
+  ],
+  "seeds": [
+    0,
+    1,
+    2
+  ],
+  "training": {
+    "learning_rate": 0.001,
+    "weight_decay": 0.0001,
+    "batch_size": 512,
+    "max_epochs": 80,
+    "patience": 10,
+    "min_delta": 0.0001
+  },
+  "hidden": {
+    "resnet18": {
+      "local": 128,
+      "MEAN": 128,
+      "GCN": 128,
+      "SAGE": 67
+    },
+    "phikon": {
+      "local": 128,
+      "MEAN": 128,
+      "GCN": 128,
+      "SAGE": 66
+    }
+  },
+  "dropout": 0.1,
+  "source_fit": [
+    "NCBI783",
+    "NCBI785"
+  ],
+  "validation": "TENX95",
+  "development": "TENX99",
+  "external": [
+    "1160920F",
+    "CID4290",
+    "CID44971",
+    "BASS_P01",
+    "BASS_P02",
+    "BASS_P03",
+    "BASS_P04",
+    "BASS_P05",
+    "BASS_P06",
+    "BASS_P07",
+    "BASS_P08",
+    "BASS_P09",
+    "BASS_P11",
+    "BASS_P12",
+    "BASS_P13",
+    "BASS_P14",
+    "BASS_P15",
+    "BASS_P17",
+    "BASS_P18",
+    "BASS_P19",
+    "BASS_P20",
+    "BASS_P21",
+    "BASS_P22",
+    "BASS_P23",
+    "BASS_P24"
+  ],
+  "encoder_revision": "057cc0295895c2df3dd7681a89680da6015cbefe",
+  "encoder_sha256": "d01642bfeb683c6b039622eca273e0a8e4208065a4595b63edba178c1c4cfa68",
+  "encoder_threads": 1,
+  "head_threads": 4,
+  "encoder_batch_size": 8,
+  "encoder_max_wall_hours": 6.5,
+  "head_max_wall_hours": 1,
+  "evaluation_max_wall_hours": 1,
+  "max_rss_gib": 10,
+  "primary_budget": 0.2,
+  "budgets": [
+    0.1,
+    0.2,
+    0.4
+  ],
+  "mask_seeds": [
+    0,
+    1,
+    2
+  ],
+  "bootstrap_resamples": 10000,
+  "bootstrap_seed": 20261101,
+  "serialization_rtol": 0,
+  "serialization_atol": 0,
+  "leakage_rtol": 0,
+  "leakage_atol": 0,
+  "minimum_shuffled_changed_edge_fraction": 0.5,
+  "encoder_workers": 4,
+  "training_refinement": {
+    "version": 2,
+    "learning_rates": [
+      0.001,
+      0.0001
+    ],
+    "selection": "Mean best qualifying TENX95 MSE across all three seeds, separately per backbone/model/control; exact tie chooses smaller learning rate.",
+    "external_scoring": "Only validation-selected learning rate; all three seeds. Existing complete ResNet1e-3 exports remain a separately labelled sensitivity.",
+    "other_settings_unchanged": true,
+    "refinement_reason": "17 of18 first Phikon fits selected epoch1; validation deterioration with declining fitting loss. No Phikon development/external predictions or outcomes have been inspected."
+  }
+}