Add GSE194040 pCR expression Elastic Net model

.gitignore

@@ -599,4 +599,10 @@ yarn-error.log
 
 # End of https://www.toptal.com/developers/gitignore/api/python,node,vue,vuejs,matlab
 
+# Include only the submitted GSE194040 package provenance logs.
+!flask_backend/models/gse194040_pcr_expression_elastic_net/logs/
+flask_backend/models/gse194040_pcr_expression_elastic_net/logs/*
+!flask_backend/models/gse194040_pcr_expression_elastic_net/logs/gse194040_model_training_session_info.json
+!flask_backend/models/gse194040_pcr_expression_elastic_net/logs/gse194040_session_info.json
+
 

flask_backend/models/gse194040_pcr_expression_elastic_net/.dockerignore

@@ -0,0 +1,16 @@
+.git
+.gitignore
+.venv
+venv
+env
+ENV
+__pycache__
+*.py[cod]
+.pytest_cache
+.coverage
+htmlcov
+tmp
+temp
+*.tmp
+*.pptx
+*.pdf

flask_backend/models/gse194040_pcr_expression_elastic_net/CITATION.cff

@@ -0,0 +1,37 @@
+cff-version: 1.2.0
+message: >-
+  If you use this package, cite this software record and the source
+  GSE194040 dataset.
+title: "GSE194040 exploratory pooled pCR expression-classifier pipeline"
+type: software
+authors:
+  - family-names: "Charles"
+    given-names: "Rhea"
+    affiliation: "George Washington University HIVE Lab"
+    email: "riocx1997@gmail.com"
+abstract: >-
+  A reproducible research package that prepares the public GSE194040/I-SPY2
+  gene-expression data and applies a frozen Elastic Net logistic-regression
+  pipeline to predict pathological complete response.
+keywords:
+  - GSE194040
+  - I-SPY2
+  - breast cancer
+  - gene expression
+  - elastic net
+  - machine learning
+references:
+  - type: data
+    title: "I-SPY2-990 mRNA/RPPA Data Resource: mRNA component"
+    authors:
+      - family-names: "Wolf"
+        given-names: "Denise"
+      - family-names: "Yau"
+        given-names: "Christina"
+      - family-names: "van 't Veer"
+        given-names: "Laura"
+    identifiers:
+      - type: other
+        value: "GSE194040"
+        description: "NCBI Gene Expression Omnibus accession"
+    url: "https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE194040"

flask_backend/models/gse194040_pcr_expression_elastic_net/Dockerfile

@@ -0,0 +1,18 @@
+FROM python:3.12.13-slim
+
+LABEL org.opencontainers.image.title="GSE194040 exploratory pCR expression classifier"
+LABEL org.opencontainers.image.description="Reproducible research package for the pooled GSE194040 Elastic Net pCR classifier"
+
+ENV PYTHONDONTWRITEBYTECODE=1 \
+    PYTHONUNBUFFERED=1 \
+    PIP_NO_CACHE_DIR=1
+
+WORKDIR /app
+
+COPY requirements-lock.txt .
+RUN python -m pip install --upgrade pip==25.0.1 \
+    && python -m pip install -r requirements-lock.txt
+
+COPY . .
+
+CMD ["python", "scripts/predict_gse194040_single_patient.py", "--help"]

flask_backend/models/gse194040_pcr_expression_elastic_net/LICENSES.md

@@ -0,0 +1,92 @@
+# Licensing status and scope
+
+This file records only licence information verified inside this package. It is
+not itself a licence grant, and one component's licence must not be assumed to
+cover another component.
+
+## Package contributor metadata
+
+Rhea Charles is the sole package contributor, affiliated with George Washington
+University HIVE Lab, with the role PredictMod Volunteer and contact email
+`riocx1997@gmail.com`. No ORCID was provided. This contributor record does not
+assign or alter any licence terms.
+
+## PredictMod repository code
+
+The upstream `GW-HIVE/PredictMod` repository code is licensed under the MIT
+License, with the copyright notice:
+
+`Copyright © 2022 HIVE-Lab`
+
+Source: the upstream `GW-HIVE/PredictMod` `LICENSE` file. This verified
+repository-level licence does not by itself assign a licence to the submitted
+package code or trained model.
+
+## Submitted trained model
+
+The submitted trained model is licensed under the Creative Commons Attribution
+4.0 International licence (CC BY 4.0), based on the PredictMod User Guide's
+statement that models shared through PredictMod are available to users and
+commercial entities under CC BY 4.0.
+
+References:
+
+- HIVE PredictMod User Guide:
+  `https://hivelab.biochemistry.gwu.edu/predictmod/help`
+- Upstream user-guide source:
+  `https://github.com/GW-HIVE/PredictMod/blob/main/vuetify_app/src/docs/help_content/help.md`
+- CC BY 4.0 terms: `https://creativecommons.org/licenses/by/4.0/`
+
+This model licence does not assign CC BY 4.0 to the original GSE194040 source
+data.
+
+## BioCompute Object
+
+`bco/gse194040_pcr_pipeline_bco_v1.json` records
+`https://spdx.org/licenses/CC-BY-4.0.html` in its provenance domain. This is the
+verified licence statement for the BCO document itself. It does not assign that
+licence to the repository code, submitted model, source data, or vendored
+schema snapshot.
+
+## GSE194040 data
+
+The package attributes its source data to the NCBI Gene Expression Omnibus
+accession `GSE194040`, titled "I-SPY2-990 mRNA/RPPA Data Resource: mRNA
+component." The GEO record lists Denise Wolf, Christina Yau, and Laura
+van 't Veer as series contributors and associate PubMed identifiers `35623341`
+and `40905691` with the series.
+
+The downloaded GEO archives and the large processed training CSV are not
+distributed in this package. Original GSE194040 source data are governed by
+NCBI GEO and the original data submitters' terms. Neither the PredictMod MIT
+software licence nor PredictMod's CC BY 4.0 model/database-set terms are
+asserted to cover the original GEO data.
+
+Source record:
+`https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE194040`
+
+## PredictMod database sets
+
+PredictMod applies CC BY 4.0 to its database sets, based on the HIVE PredictMod
+website licence statement. This scope is distinct from both repository source
+code and original third-party source data.
+
+References:
+
+- HIVE PredictMod website:
+  `https://hivelab.biochemistry.gwu.edu/predictmod/`
+- Upstream website licence component:
+  `https://github.com/GW-HIVE/PredictMod/blob/main/vuetify_app/src/views/LicenseShow.vue`
+- CC BY 4.0 terms: `https://creativecommons.org/licenses/by/4.0/`
+
+## Vendored IEEE 2791 schema snapshot
+
+`schemas/ieee-2791-v1.4/README.md` records the source and retrieval date of the
+vendored schema files, but the package contains no licence or redistribution
+terms for that snapshot. An authorized maintainer must confirm that the schema
+files may be redistributed and add any required notice.
+
+## Information still required
+
+Before publication or redistribution, confirm the licence and notice
+requirements for the vendored IEEE 2791 schema snapshot.

flask_backend/models/gse194040_pcr_expression_elastic_net/PACKAGE_MANIFEST.md

@@ -0,0 +1,89 @@
+# Package Manifest
+
+The submission package contains 120 files. The full processed training CSV,
+downloaded GEO archives, Git LFS rules, generated Python and pytest caches,
+volunteership reporting, symposium material, and package-audit documentation
+are intentionally excluded.
+
+## Top-level inventory
+
+| Path | File count | Contents |
+|---|---:|---|
+| `bco/` | 1 | IEEE 2791 BioCompute Object |
+| `data_processed/` | 7 | Data dictionaries, manifests, and preprocessing records |
+| `docs/` | 7 | Technical model and analysis documentation |
+| `examples/` | 1 | Single-patient example input |
+| `logs/` | 2 | Preprocessing and model-training session information |
+| `models/` | 7 | Frozen model pipelines and selected-model metadata |
+| `results/` | 60 | Tables, figures, evaluations, and validation records |
+| `schemas/` | 9 | Vendored IEEE 2791 v1.4 schemas |
+| `scripts/` | 9 | Preprocessing, training, evaluation, validation, and prediction code |
+| `tests/` | 6 | Automated test modules |
+| Package-root files | 11 | README, manifest, citation, licensing, dependency, Docker, and pytest configuration |
+
+## Required completion-criteria files
+
+| Requirement | Included path |
+|---|---|
+| Historical training-dataset provenance | `data_processed/gse194040_preprocessing_summary.md` and `results/reproducibility/gse194040_reproducibility_manifest_v1_20260719.json` |
+| Model training | `scripts/train_gse194040_models.py` |
+| Single-patient prediction | `scripts/predict_gse194040_single_patient.py` |
+| Example patient input | `examples/gse194040_single_patient_input_example.csv` |
+| Example prediction output | `results/tables/gse194040_single_patient_prediction_example_v2_uncalibrated_score.csv` |
+| Frozen selected model | `models/gse194040_best_model.joblib` |
+| BioCompute Object | `bco/gse194040_pcr_pipeline_bco_v1.json` |
+
+## Reproducibility and deployment
+
+- Citation metadata: `CITATION.cff`
+- Licence scope and unresolved terms: `LICENSES.md`
+- Third-party inventory: `THIRD_PARTY_NOTICES.md`
+- Exact Python packages: `requirements-lock.txt`
+- Direct dependencies: `requirements.txt`
+- Conda environment: `environment.yml`
+- Docker build: `Dockerfile`
+- Docker exclusions: `.dockerignore`
+- Large-data policy: processed training data and downloaded GEO archives are recreated with `scripts/preprocess_gse194040.py`; no package Git LFS rules remain
+- Automated tests: `tests/`
+- IEEE 2791 v1.4 schemas: `schemas/ieee-2791-v1.4/`
+- BCO validation: `scripts/validate_gse194040_bco.py`
+
+## Outputs
+
+- Canonical model summary: `results/gse194040_model_training_summary.md`
+- Tables: `results/tables/`
+- Figures: `results/figures/`
+- Locked evaluation and confidence intervals: `results/evaluation/`
+- Calibration diagnostics: `results/calibration/`
+- Reproducibility manifest: `results/reproducibility/`
+- Test and BCO validation records: `results/validation/`
+
+## Included technical documentation
+
+- `docs/gse194040_bco_pipeline_documentation.md`
+- `docs/gse194040_calibration_analysis.md`
+- `docs/gse194040_clinical_model_comparison.md`
+- `docs/gse194040_preprocessing_leakage_audit.md`
+- `docs/gse194040_reproducibility.md`
+- `docs/gse194040_single_patient_prediction_usage.md`
+- `docs/gse194040_subgroup_evaluation.md`
+
+## Intentional exclusions
+
+- `.pytest_cache/`
+- `scripts/__pycache__/`
+- `tests/__pycache__/`
+- `docs/gse194040_volunteership_report.md`
+- `docs/gse194040_symposium_slide_outline.md`
+- `docs/gse194040_final_project_verification.md`
+- `docs/gse194040_package_audit.md`
+- `data_processed/gse194040_ml_ready_pcr_expression.csv`
+- `GSE194040/`
+- `.gitattributes`
+
+## Scope
+
+This package contains the canonical pooled full-cohort model. The separate
+Paclitaxel-only analysis is not a replacement production model and is excluded
+pending a project-lead decision on whether it should be a documented
+supplement.