Add --LR_model and --demo options to gene_vec_model.py

[chronicgiardia]

Summary

Adds two optional CLI arguments to src/gene_vec_model.py and documents them:

• --LR_model <dir>: after training gene embeddings, trains (or loads, if already present) a logistic regression model that predicts whether a pair of genes is associated, using the element-wise product of the two genes' embeddings as features. Persisted as gene_pair_lr.pkl in the directory. Requires scikit-learn.
• --demo <dir>: exports a copy of the learned embeddings into the demo directory (into its data/ subdirectory when present) so the demo scripts can consume freshly trained embeddings.

To make this testable, the training pipeline was refactored into a main() guarded by if __name__ == '__main__', and the new logic was extracted into small helpers (build_pair_features, train_or_load_lr_model, export_embeddings_for_demo). CLI behavior is unchanged.

Tests

Adds src/test_gene_vec_model.py (stdlib unittest, no pytest required) covering argument parsing, feature construction, LR model save/load (including load-without-retrain), and demo export. Run with:

cd src && python3 -m unittest test_gene_vec_model -v

All 11 tests pass.

Conversation: https://app.warp.dev/conversation/2180fe28-77cf-4cc1-9147-26b5fd7a2e6a

[Copilot]

Pull request overview

This PR extends the src/gene_vec_model.py training script with two optional CLI flags to (1) train/load a downstream logistic regression model on learned gene embeddings and (2) export trained embeddings into a demo directory, and adds unit tests and README documentation for these behaviors.

Changes:

• Added --LR_model <dir> and helper logic to train or load a persisted gene-pair logistic regression model.
• Added --demo <dir> and helper logic to export the embeddings CSV into the demo directory (preferring demo/data/ when present).
• Refactored the script to a main() entrypoint guarded by if __name__ == '__main__' and introduced unittest coverage for the new helpers/CLI parsing.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
src/gene_vec_model.py	Adds new CLI args, extracts helper functions, and wraps training flow into a main() entrypoint; optionally trains/loads LR and exports embeddings.
src/test_gene_vec_model.py	Adds unittest coverage for argument parsing, feature construction, LR save/load behavior, and demo export (skips LR tests if sklearn missing).
README.md	Documents the new --LR_model and --demo options and their outputs/behavior.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[chronicgiardia]

Updated: replaced the argparse-based parser in src/gene_vec_model.py with a manual sys.argv scanner (explicit per-flag cursor loop supporting both --flag value and --flag=value forms, manual --help, and error+exit(1) on unrecognized arguments). Behavior and the args.* attribute interface are unchanged; all 11 existing tests pass.