Add GSE194040 pCR expression Elastic Net model

[riocx978]

Summary

This pull request adds a gene-expression model for predicting pathological complete response (pCR) using the GSE194040 I-SPY2 breast-cancer dataset.

The submitted frozen model is an Elastic Net logistic-regression pipeline using 19,134 gene-expression features.

The model was developed using 987 patients pooled across 10 neoadjuvant treatment arms. It predicts pCR within this mixed-treatment population and is not intended to select treatments or serve as clinical decision support.

Included

• Frozen trained Elastic Net model
• Single-patient prediction script
• Example input and expected output
• Training and preprocessing scripts
• BioCompute Object and validation record
• Model metadata and feature manifest
• Evaluation, calibration and subgroup results
• Reproducibility documentation and automated tests

Validation

• 41 automated tests passed
• BCO schema validation passed with zero errors
• Frozen model SHA-256:
  e3bbccac97c4495dbb8e8c66a8f35921ca6ff20cbb9e0b0671aaf0a00504f9d4
• Expected prediction SHA-256:
  b17b9dc570b92f3efb6fce2548134c1e9fd48a890f5c4d3ddfec6648fcf0523c
• Final package size: approximately 16.3 MiB
• No Git LFS files are required

Data availability

The large processed training matrix and downloaded GEO archives are not included.

The source dataset is publicly available through NCBI GEO under accession GSE194040. The package includes preprocessing scripts, checksums, manifests and documentation needed to reproduce the processed dataset.

Licensing

• PredictMod repository source code: MIT
• Submitted trained model: CC BY 4.0
• BioCompute Object: CC BY 4.0
• Original GSE194040 data remains subject to NCBI GEO and the original submitters’ terms