How srini911/tcga-brca-subtype-classifier works
This repository is a reproducible research workflow for classifying breast tumors into four PAM50 subtypes from TCGA RNA-seq. Its principal path downloads cohort data, transforms and labels expression features, compares classifiers, and produces evaluation results. Two companion analyses inspect influential genes and subtype confusions, and test smaller gene panels. The graph follows the documented script order; source excerpts confirm the major data reads, model work, and result writes. Raw and processed data directories are generated and gitignored, so they are not shown as repository files.
Main entry points: 01_download.py Download cohort · 04_interpret.py Interpret models · 05_gene_panel.py Test gene panels · 03_train_evaluate.py Train and evaluate
Components 12
Data preparation
- Download cohort
src/01_download.py - Prepare features
src/02_preprocess.py
Model evaluation
- Train and evaluate
src/03_train_evaluate.py - Evaluation results
results/metrics.json
Model analysis
- Interpret models
src/04_interpret.py - Test gene panels
src/05_gene_panel.py - Gene panel results
results/panel_top50_genes.csv - Ranked gene results
results/top_genes.csv
Shared utilities
- Project settings
src/config.py - Data and figure helpers
src/utils.py
Other components
- Researcher
- cBioPortal datahub
Connections 22
- Researcher → Download cohort: starts workflow
README.md - Download cohort → cBioPortal datahub: downloads cohort files
src/01_download.py - Download cohort → Prepare features: provides raw inputs
README.md - Prepare features → Train and evaluate: provides processed features
README.md - Researcher → Train and evaluate: runs evaluation
README.md - Train and evaluate → Evaluation results: writes metrics
src/03_train_evaluate.py - Researcher → Interpret models: runs interpretation
README.md - Interpret models → Evaluation results: reads and updates
src/04_interpret.py - Interpret models → Ranked gene results: writes rankings
src/04_interpret.py - Researcher → Test gene panels: runs panel experiment
README.md - Test gene panels → Gene panel results: writes top-50 list
src/05_gene_panel.py - Download cohort → Project settings: uses settings
src/01_download.py - Prepare features → Project settings: uses settings
src/02_preprocess.py - Train and evaluate → Project settings: uses settings
src/03_train_evaluate.py - Interpret models → Project settings: uses settings
src/04_interpret.py - Test gene panels → Project settings: uses settings
src/05_gene_panel.py - Download cohort → Data and figure helpers: sets logging
src/01_download.py - Prepare features → Data and figure helpers: sets logging
src/02_preprocess.py - Train and evaluate → Data and figure helpers: loads data and saves figures
src/03_train_evaluate.py - Interpret models → Data and figure helpers: loads data and saves figures
src/04_interpret.py - Test gene panels → Data and figure helpers: loads data and saves figures
src/05_gene_panel.py - Test gene panels → Gene panel results: writes panel metrics
src/05_gene_panel.py