Biostats Final Project

Env

• create conda env with conda env create --prefix ./env -f environment.yaml
• activate with conda activate ./env

Data

• Download files Model.csv, OmicsExpressionTPMLogp1HumanProteinCodingGenes.csv, OmicsProfiles.csv from https://depmap.org/portal/data_page/?tab=allData and place in ./data/raw
• create folds with below
• will create 5 folds

python .\src\create_folds.py -X .\data\raw\OmicsExpressionTPMLogp1HumanProteinCodingGenes.csv -y .\data\raw\OmicsProfiles.csv

Train

• model details can be found in ./src/model.py
• Train with python ./src/train.py

Evaluation

• calculate test accuracies and create plots with python ./src/test_performance.py
• calculate pearson correlation between performance metrics with python ./src/calc_correlations.py
• calculate shap values and create feature importance plot with python ./src/shap_reports.py

Report

Our final report can be found under final_report.pdf.

Figures and related code:

• Figures 1-4: ./figures/Figures_1-4_Gabe.ipynb
• Figure 5: ./reports/performance/accuracy_by_fold.png
• Figure 6: ./reports/performance/accuracy_by_lineage.png
• Figure 7: ./reports/shap/aggregate_feature_importance.png
• Figure 8: ./figures/figure_8_10_kristi.ipynb
• Figure 9: ./figures/figs_9_11-13_morgan.ipynb
• Figure 10: ./figures/figure_8_10_kristi.ipynb
• Figure 11-13: ./figures/figs_9_11-13_morgan.ipynb