pooledscreenid 0.1.1__tar.gz

pooledscreenid-0.1.1/.github/workflows/tests.yml

@@ -0,0 +1,20 @@
+name: tests
+
+on:
+  push:
+  pull_request:
+
+jobs:
+  pytest:
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        python-version: ["3.10", "3.11", "3.12", "3.13"]
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+      - run: python -m pip install --upgrade pip
+      - run: python -m pip install -e . pytest
+      - run: python -m pytest -q

pooledscreenid-0.1.1/CHANGELOG.md

@@ -0,0 +1,17 @@
+# Changelog
+
+## 0.1.1 - 2026-08-29
+
+- Added `external_calibration` as the fourth independent evidence-vector axis.
+- Added `resolution_inputs` so the summary claim resolution is auditable from
+  the four source fields.
+- Upgraded the claim-ledger schema to 1.1.0.
+- Added schema-focused tests and a Python 3.10-3.13 continuous-integration
+  matrix.
+- Clarified that this article-linked release follows semantic versioning and
+  has no third-party runtime dependencies.
+
+## 0.1.0 - 2026-08-29
+
+- First public release with Python API, command-line interface, frozen EGFR and
+  MET examples, integrity manifests and machine-readable claim output.

pooledscreenid-0.1.1/CITATION.cff

@@ -0,0 +1,20 @@
+cff-version: 1.2.0
+message: "If you use PooledScreenID, cite the archived software release and the accompanying methods article."
+title: "PooledScreenID: outcome-blind shortcut and separability diagnostics for pooled variant-effect screens"
+type: software
+version: 0.1.1
+date-released: 2026-08-29
+license: MIT
+authors:
+  - family-names: Niu
+    given-names: Niu
+  - family-names: Wei
+    given-names: Fang
+  - family-names: Wang
+    given-names: Yan
+  - family-names: Liu
+    given-names: Hu
+  - family-names: Wu
+    given-names: Bin
+repository-code: "https://doi.org/10.5281/zenodo.22156510"
+doi: "10.5281/zenodo.22156510"

pooledscreenid-0.1.1/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 PooledScreenID contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

pooledscreenid-0.1.1/MANIFEST.in

@@ -0,0 +1,4 @@
+include README.md LICENSE CITATION.cff THIRD_PARTY_NOTICES.md CHANGELOG.md
+recursive-include examples *.json *.csv *.py *.md
+recursive-include schemas *.json
+recursive-include .github *.yml

pooledscreenid-0.1.1/PKG-INFO

@@ -0,0 +1,109 @@
+Metadata-Version: 2.4
+Name: pooledscreenid
+Version: 0.1.1
+Summary: Outcome-blind shortcut and separability diagnostics for pooled variant-effect screens
+Author: Niu Niu, Fang Wei, Yan Wang, Hu Liu, Bin Wu
+License-Expression: MIT
+Project-URL: Documentation, https://doi.org/10.5281/zenodo.22156510
+Project-URL: Repository, https://doi.org/10.5281/zenodo.22156510
+Keywords: deep mutational scanning,cold start,negative control,identifiability,variant effect
+Classifier: Development Status :: 4 - Beta
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Scientific/Engineering :: Bio-Informatics
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Dynamic: license-file
+
+# PooledScreenID
+
+PooledScreenID is a small, installable Python package for reporting four
+quantities separately in cold-start pooled variant-effect analyses:
+
+1. material predictive gain;
+2. association with an outcome-blind control channel;
+3. representation separability at a prespecified operating scale;
+4. external calibration, kept distinct from the target estimand; and
+5. the resulting allowed, restricted and prohibited claims.
+
+The package does not infer a molecular mechanism and does not make treatment
+recommendations. A cold split is treated as a transport test, not as proof of
+the information source used by a model.
+
+## Install
+
+```bash
+python -m pip install pooledscreenid-0.1.1-py3-none-any.whl
+```
+
+For a source checkout:
+
+```bash
+python -m pip install -e .
+```
+
+## Minimal Python API
+
+```python
+from pooledscreenid import run_audit
+
+report, ledger = run_audit(
+    config_path="examples/configs/egfr.json",
+    output_dir="example_output/egfr",
+)
+print(report["claim_resolution"])
+```
+
+## Command line
+
+```bash
+pooledscreenid audit \
+  --config examples/configs/egfr.json \
+  --output-dir example_output/egfr
+
+pooledscreenid audit \
+  --config examples/configs/met.json \
+  --output-dir example_output/met
+```
+
+Each run writes `report.json`, `report.md` and `claim_ledger.json`. Inputs,
+configuration, source hashes and expected outputs are frozen in the release.
+The ledger exposes `material_signal`, `control_association`, `separability` and
+`external_calibration` as independent fields. `claim_resolution` is generated
+from those fields under priority-ordered rules and never replaces them.
+
+PooledScreenID 0.1.1 is the article-linked frozen release. The public API is
+managed with semantic versioning. It supports Python 3.10-3.13 and has no
+third-party runtime dependencies; the test suite uses pytest.
+
+## Included end-to-end examples
+
+- `EGFR`: ten frozen variant-cold folds from the L858R-background pooled EGFR
+  inhibitor screen, plus the outcome-free geometry and conditional operating
+  boundary used in the manuscript.
+- `MET`: ten frozen variant-cold folds from an independently selected pooled
+  MET inhibitor screen. The example demonstrates a result below the materiality
+  gate and explicitly leaves separability unevaluated.
+
+These examples begin with frozen, source-derived fold diagnostics rather than
+raw sequencing reads. They reproduce the evidence-to-claim stage of the
+framework; raw count processing and model fitting remain documented in the
+accompanying analysis repository.
+
+## Interpretation order
+
+The software reports the four-axis evidence vector before returning a summary label. If
+shared predictability exceeds its prespecified boundary, or lies inside a
+prespecified boundary zone, fine-grained attribution is restricted before any
+control-specific contrast is interpreted. Absence of a material increment is
+not an equivalence or biological-null claim.
+
+## Release integrity
+
+`RELEASE_MANIFEST.json` and `SHA256SUMS.txt` bind the software, tests, continuous-integration configuration, examples,
+configuration files and claim ledger to this release. The Zenodo record is the
+permanent release series: <https://doi.org/10.5281/zenodo.22156510>.

pooledscreenid-0.1.1/README.md

@@ -0,0 +1,88 @@
+# PooledScreenID
+
+PooledScreenID is a small, installable Python package for reporting four
+quantities separately in cold-start pooled variant-effect analyses:
+
+1. material predictive gain;
+2. association with an outcome-blind control channel;
+3. representation separability at a prespecified operating scale;
+4. external calibration, kept distinct from the target estimand; and
+5. the resulting allowed, restricted and prohibited claims.
+
+The package does not infer a molecular mechanism and does not make treatment
+recommendations. A cold split is treated as a transport test, not as proof of
+the information source used by a model.
+
+## Install
+
+```bash
+python -m pip install pooledscreenid-0.1.1-py3-none-any.whl
+```
+
+For a source checkout:
+
+```bash
+python -m pip install -e .
+```
+
+## Minimal Python API
+
+```python
+from pooledscreenid import run_audit
+
+report, ledger = run_audit(
+    config_path="examples/configs/egfr.json",
+    output_dir="example_output/egfr",
+)
+print(report["claim_resolution"])
+```
+
+## Command line
+
+```bash
+pooledscreenid audit \
+  --config examples/configs/egfr.json \
+  --output-dir example_output/egfr
+
+pooledscreenid audit \
+  --config examples/configs/met.json \
+  --output-dir example_output/met
+```
+
+Each run writes `report.json`, `report.md` and `claim_ledger.json`. Inputs,
+configuration, source hashes and expected outputs are frozen in the release.
+The ledger exposes `material_signal`, `control_association`, `separability` and
+`external_calibration` as independent fields. `claim_resolution` is generated
+from those fields under priority-ordered rules and never replaces them.
+
+PooledScreenID 0.1.1 is the article-linked frozen release. The public API is
+managed with semantic versioning. It supports Python 3.10-3.13 and has no
+third-party runtime dependencies; the test suite uses pytest.
+
+## Included end-to-end examples
+
+- `EGFR`: ten frozen variant-cold folds from the L858R-background pooled EGFR
+  inhibitor screen, plus the outcome-free geometry and conditional operating
+  boundary used in the manuscript.
+- `MET`: ten frozen variant-cold folds from an independently selected pooled
+  MET inhibitor screen. The example demonstrates a result below the materiality
+  gate and explicitly leaves separability unevaluated.
+
+These examples begin with frozen, source-derived fold diagnostics rather than
+raw sequencing reads. They reproduce the evidence-to-claim stage of the
+framework; raw count processing and model fitting remain documented in the
+accompanying analysis repository.
+
+## Interpretation order
+
+The software reports the four-axis evidence vector before returning a summary label. If
+shared predictability exceeds its prespecified boundary, or lies inside a
+prespecified boundary zone, fine-grained attribution is restricted before any
+control-specific contrast is interpreted. Absence of a material increment is
+not an equivalence or biological-null claim.
+
+## Release integrity
+
+`RELEASE_MANIFEST.json` and `SHA256SUMS.txt` bind the software, tests, continuous-integration configuration, examples,
+configuration files and claim ledger to this release. The Zenodo record is the
+permanent release series: <https://doi.org/10.5281/zenodo.22156510>.

pooledscreenid-0.1.1/THIRD_PARTY_NOTICES.md

@@ -0,0 +1,16 @@
+# Third-party data notices
+
+The release contains only small, derived fold-level diagnostic tables and
+machine-readable summaries required to replay the published examples. It does
+not redistribute raw GSE305057 sequencing files, the raw MET screen, article
+PDFs, clinical records, patient-level data or licensed databases.
+
+- EGFR example: derived from NCBI GEO accession GSE305057. Users should cite
+  the source study and retrieve raw data from GEO under the source terms.
+- MET example: derived from the public Frase laboratory MET kinase inhibitor
+  DMS repository. Users should cite the source repository and study.
+- Robichaux independent-construct results are described in the manuscript but
+  are not redistributed in this minimal software release.
+
+Derived tables are provided for computational reproducibility and retain
+source identifiers in their configuration and provenance records.

pooledscreenid-0.1.1/examples/configs/egfr.json

@@ -0,0 +1,29 @@
+{
+  "dataset_id": "EGFR_GSE305057",
+  "fold_gains_csv": "../data/egfr_fold_gains.csv",
+  "materiality_threshold": 0.02,
+  "target_representation": "smooth residue-position representation",
+  "outcome_blind_control": "DMSO count/QC atlas",
+  "prediction_unit": "variant-condition edge",
+  "independent_unit": "unique variant with position-aware resampling",
+  "block_unit": "20-residue library block",
+  "estimand": "variant-cold MAE gain of position and its residual gain after the outcome-blind control atlas",
+  "eta": 0.21321233482512197,
+  "eta_boundary": 0.219,
+  "boundary_zone": 0.01,
+  "source_id": "NCBI GEO GSE305057; L858R-background second-site pooled screen",
+  "external_calibration": {
+    "status": "supportive_distinct_estimand",
+    "source": "Robichaux et al. independent Ba/F3 construct panel",
+    "estimand": "cold-start prediction from the source study's coarse structure-group labels",
+    "note": "Positive calibration at a distinct coarse grouping estimand; it does not validate fine-grained continuous structure in GSE305057."
+  },
+  "expected_claim_resolution": "restricted_boundary_proximal",
+  "metadata": {
+    "position_unique_variance_fraction_best_direction": 0.05743890584691004,
+    "continuous_structure_gain_after_control_and_position": -0.002720124638247945,
+    "control_atlas_unconditional_gain": 0.10628083240037886,
+    "control_given_position_gain": 0.00844304967707693,
+    "interpretation": "control-associated positional result; fine-grained attribution restricted"
+  }
+}

pooledscreenid-0.1.1/examples/configs/met.json

@@ -0,0 +1,27 @@
+{
+  "dataset_id": "MET_PUBLIC_DMS",
+  "fold_gains_csv": "../data/met_fold_gains.csv",
+  "materiality_threshold": 0.02,
+  "target_representation": "smooth residue-position representation",
+  "outcome_blind_control": "released DMSO fitness/SD covariates (Tier B control)",
+  "prediction_unit": "variant-inhibitor edge",
+  "independent_unit": "unique variant",
+  "block_unit": "20-residue block",
+  "estimand": "variant-cold MAE gain of position and residual gain after released control covariates",
+  "eta": null,
+  "eta_boundary": null,
+  "boundary_zone": 0.0,
+  "source_id": "Frase laboratory public MET kinase inhibitor DMS repository",
+  "external_calibration": {
+    "status": "external_contrast_only",
+    "source": "Independent MET pooled inhibitor screen",
+    "estimand": "cross-dataset contrast in shortcut magnitude",
+    "note": "The MET analysis is an external contrast, not calibration of the EGFR target estimand."
+  },
+  "expected_claim_resolution": "no_material_increment",
+  "metadata": {
+    "block_permutation_null_mean": 0.006022973402010649,
+    "block_permutation_p": 0.005988023952095809,
+    "interpretation": "no EGFR-sized material position gain; smaller shortcuts not excluded"
+  }
+}

pooledscreenid-0.1.1/examples/data/egfr_fold_gains.csv

@@ -0,0 +1,11 @@
+fold,unadjusted_gain,adjusted_gain
+1,0.1059777304912998,0.0048186999161899
+2,0.0905323124916878,0.0004236617897617
+3,0.1124205997787652,0.0042290727688932
+4,0.0862271643262527,0.0056428031648455
+5,0.0870414775800483,0.0038583840936432
+6,0.118258013241681,0.0016648731016825
+7,0.1088183403032501,0.0017212395905227
+8,0.0972533535610891,0.0019253213041058
+9,0.0940889871896792,-0.0002014812743086
+10,0.1003905456728689,-0.0014518770517326

pooledscreenid-0.1.1/examples/data/met_fold_gains.csv

@@ -0,0 +1,11 @@
+fold,unadjusted_gain,adjusted_gain
+1,0.00963338634441313,-0.0005106063266484551
+2,0.004941608828022082,-0.0001186125773013913
+3,0.002962998695935859,4.962424146637279e-05
+4,0.004453951625545294,7.340498336572132e-05
+5,0.007491007099538871,0.00048685587075725234
+6,0.006174244928453199,0.0005966772424034206
+7,0.00642193819895609,-0.0005431709212077029
+8,0.001985832625287287,-0.00033610057399025006
+9,0.00976383637397693,-0.0004701601295301616
+10,0.008515505195033635,-1.8866541230400102e-05

pooledscreenid-0.1.1/examples/outputs/egfr/claim_ledger.json

@@ -0,0 +1,92 @@
+{
+  "schema_version": "1.1.0",
+  "software": {
+    "name": "pooledscreenid",
+    "version": "0.1.1"
+  },
+  "dataset_id": "EGFR_GSE305057",
+  "evidence": [
+    {
+      "evidence_id": "EGFR_GSE305057:screen-audit-v1",
+      "material_signal": "supported",
+      "control_association": "supported",
+      "separability": "boundary_proximal",
+      "external_calibration": {
+        "status": "supportive_distinct_estimand",
+        "source": "Robichaux et al. independent Ba/F3 construct panel",
+        "estimand": "cold-start prediction from the source study's coarse structure-group labels",
+        "note": "Positive calibration at a distinct coarse grouping estimand; it does not validate fine-grained continuous structure in GSE305057."
+      },
+      "claim_resolution": "restricted_boundary_proximal",
+      "resolution_inputs": {
+        "material_signal": "supported",
+        "control_association": "supported",
+        "separability": "boundary_proximal",
+        "external_calibration": "supportive_distinct_estimand"
+      },
+      "config_sha256": "bdad818be7f2a25a861011558d3dc88678007875b027eb9f0498f181efe7bded",
+      "input_sha256": "01f4921e8b7238ddb68717c87fc7ed7d7c12c85686ff25f70cdfd1b1d1a361ef"
+    }
+  ],
+  "allowed_claims": [
+    {
+      "text": "The frozen variant-cold MAE gain of position and its residual gain after the outcome-blind control atlas and its uncertainty may be reported.",
+      "evidence_ids": [
+        "EGFR_GSE305057:screen-audit-v1"
+      ]
+    },
+    {
+      "text": "Cold-start transport and information-source attribution are reported separately.",
+      "evidence_ids": [
+        "EGFR_GSE305057:screen-audit-v1"
+      ]
+    }
+  ],
+  "restricted_claims": [
+    {
+      "text": "The apparent increment is associated with the outcome-blind control channel.",
+      "evidence_ids": [
+        "EGFR_GSE305057:screen-audit-v1"
+      ]
+    },
+    {
+      "text": "Fine-grained attribution is restricted by the prespecified separability assessment.",
+      "evidence_ids": [
+        "EGFR_GSE305057:screen-audit-v1"
+      ]
+    },
+    {
+      "text": "External positive calibration concerns a distinct estimand and does not validate the target representation.",
+      "evidence_ids": [
+        "EGFR_GSE305057:screen-audit-v1"
+      ]
+    }
+  ],
+  "prohibited_claims": [
+    {
+      "text": "Prediction alone does not establish a binding or causal mechanism.",
+      "evidence_ids": [
+        "EGFR_GSE305057:screen-audit-v1"
+      ]
+    },
+    {
+      "text": "This analysis does not support patient-level treatment recommendation.",
+      "evidence_ids": [
+        "EGFR_GSE305057:screen-audit-v1"
+      ]
+    },
+    {
+      "text": "Do not claim that adjustment proves the shared signal is purely technical.",
+      "evidence_ids": [
+        "EGFR_GSE305057:screen-audit-v1"
+      ]
+    },
+    {
+      "text": "Do not make a structure-specific attribution from this dataset.",
+      "evidence_ids": [
+        "EGFR_GSE305057:screen-audit-v1"
+      ]
+    }
+  ],
+  "human_review_required": true
+}

pooledscreenid-0.1.1/examples/outputs/egfr/report.json

@@ -0,0 +1,63 @@
+{
+  "dataset_id": "EGFR_GSE305057",
+  "folds": 10,
+  "unadjusted_gain_mean": 0.10010085246366221,
+  "adjusted_gain_mean": 0.00226306974036033,
+  "materiality_threshold": 0.02,
+  "material_signal": "supported",
+  "control_association": "supported",
+  "separability": "boundary_proximal",
+  "external_calibration": {
+    "status": "supportive_distinct_estimand",
+    "source": "Robichaux et al. independent Ba/F3 construct panel",
+    "estimand": "cold-start prediction from the source study's coarse structure-group labels",
+    "note": "Positive calibration at a distinct coarse grouping estimand; it does not validate fine-grained continuous structure in GSE305057."
+  },
+  "eta": 0.21321233482512197,
+  "eta_boundary": 0.219,
+  "boundary_distance": 0.005787665174878026,
+  "claim_resolution": "restricted_boundary_proximal",
+  "allowed_claims": [
+    "The frozen variant-cold MAE gain of position and its residual gain after the outcome-blind control atlas and its uncertainty may be reported.",
+    "Cold-start transport and information-source attribution are reported separately."
+  ],
+  "restricted_claims": [
+    "The apparent increment is associated with the outcome-blind control channel.",
+    "Fine-grained attribution is restricted by the prespecified separability assessment.",
+    "External positive calibration concerns a distinct estimand and does not validate the target representation."
+  ],
+  "prohibited_claims": [
+    "Prediction alone does not establish a binding or causal mechanism.",
+    "This analysis does not support patient-level treatment recommendation.",
+    "Do not claim that adjustment proves the shared signal is purely technical.",
+    "Do not make a structure-specific attribution from this dataset."
+  ],
+  "audit_context": {
+    "prediction_unit": "variant-condition edge",
+    "independent_unit": "unique variant with position-aware resampling",
+    "block_unit": "20-residue library block",
+    "target_representation": "smooth residue-position representation",
+    "outcome_blind_control": "DMSO count/QC atlas",
+    "source_id": "NCBI GEO GSE305057; L858R-background second-site pooled screen",
+    "fold_summary": {
+      "folds": 10,
+      "unadjusted_gain_mean": 0.10010085246366221,
+      "adjusted_gain_mean": 0.00226306974036033,
+      "unadjusted_gain_range": [
+        0.0862271643262527,
+        0.118258013241681
+      ],
+      "adjusted_gain_range": [
+        -0.0014518770517326,
+        0.0056428031648455
+      ],
+      "input_sha256": "01f4921e8b7238ddb68717c87fc7ed7d7c12c85686ff25f70cdfd1b1d1a361ef"
+    },
+    "folds_are_independent_biological_units": false
+  },
+  "software": {
+    "name": "pooledscreenid",
+    "version": "0.1.1"
+  },
+  "config_sha256": "bdad818be7f2a25a861011558d3dc88678007875b027eb9f0498f181efe7bded"
+}

pooledscreenid-0.1.1/examples/outputs/egfr/report.md

@@ -0,0 +1,23 @@
+# EGFR_GSE305057 PooledScreenID report
+
+- Claim resolution: `restricted_boundary_proximal`
+- Unadjusted gain: `0.100101`
+- Adjusted gain: `0.002263`
+- Materiality threshold: `0.020000`
+- Separability: `boundary_proximal`
+- External calibration: `supportive_distinct_estimand`
+
+## Allowed claims
+- The frozen variant-cold MAE gain of position and its residual gain after the outcome-blind control atlas and its uncertainty may be reported.
+- Cold-start transport and information-source attribution are reported separately.
+
+## Restricted claims
+- The apparent increment is associated with the outcome-blind control channel.
+- Fine-grained attribution is restricted by the prespecified separability assessment.
+- External positive calibration concerns a distinct estimand and does not validate the target representation.
+
+## Prohibited interpretations
+- Prediction alone does not establish a binding or causal mechanism.
+- This analysis does not support patient-level treatment recommendation.
+- Do not claim that adjustment proves the shared signal is purely technical.
+- Do not make a structure-specific attribution from this dataset.

pooledscreenid-0.1.1/examples/outputs/met/claim_ledger.json

@@ -0,0 +1,86 @@
+{
+  "schema_version": "1.1.0",
+  "software": {
+    "name": "pooledscreenid",
+    "version": "0.1.1"
+  },
+  "dataset_id": "MET_PUBLIC_DMS",
+  "evidence": [
+    {
+      "evidence_id": "MET_PUBLIC_DMS:screen-audit-v1",
+      "material_signal": "below_threshold",
+      "control_association": "not_supported",
+      "separability": "not_evaluated",
+      "external_calibration": {
+        "status": "external_contrast_only",
+        "source": "Independent MET pooled inhibitor screen",
+        "estimand": "cross-dataset contrast in shortcut magnitude",
+        "note": "The MET analysis is an external contrast, not calibration of the EGFR target estimand."
+      },
+      "claim_resolution": "no_material_increment",
+      "resolution_inputs": {
+        "material_signal": "below_threshold",
+        "control_association": "not_supported",
+        "separability": "not_evaluated",
+        "external_calibration": "external_contrast_only"
+      },
+      "config_sha256": "64e2b9b2348b853dadff10162ae21b73905972d55bc0a37238df4b8b17084e30",
+      "input_sha256": "731660fbc4f79ed5904ae9837bece5417caf7f8e472e19ceca24e39fb82fe1c8"
+    }
+  ],
+  "allowed_claims": [
+    {
+      "text": "The frozen variant-cold MAE gain of position and residual gain after released control covariates and its uncertainty may be reported.",
+      "evidence_ids": [
+        "MET_PUBLIC_DMS:screen-audit-v1"
+      ]
+    },
+    {
+      "text": "Cold-start transport and information-source attribution are reported separately.",
+      "evidence_ids": [
+        "MET_PUBLIC_DMS:screen-audit-v1"
+      ]
+    }
+  ],
+  "restricted_claims": [
+    {
+      "text": "No material increment was observed at the frozen threshold.",
+      "evidence_ids": [
+        "MET_PUBLIC_DMS:screen-audit-v1"
+      ]
+    }
+  ],
+  "prohibited_claims": [
+    {
+      "text": "Prediction alone does not establish a binding or causal mechanism.",
+      "evidence_ids": [
+        "MET_PUBLIC_DMS:screen-audit-v1"
+      ]
+    },
+    {
+      "text": "This analysis does not support patient-level treatment recommendation.",
+      "evidence_ids": [
+        "MET_PUBLIC_DMS:screen-audit-v1"
+      ]
+    },
+    {
+      "text": "Do not interpret the result as equivalence or a biological null.",
+      "evidence_ids": [
+        "MET_PUBLIC_DMS:screen-audit-v1"
+      ]
+    },
+    {
+      "text": "Do not claim that representation separability was established.",
+      "evidence_ids": [
+        "MET_PUBLIC_DMS:screen-audit-v1"
+      ]
+    },
+    {
+      "text": "Do not claim external calibration of the target estimand.",
+      "evidence_ids": [
+        "MET_PUBLIC_DMS:screen-audit-v1"
+      ]
+    }
+  ],
+  "human_review_required": true
+}