This is the GitHub page corresponding to the paper "Multi-Modal Deep Learning-Based Model to Predict Burkitt Lymphoma Recurrence": https://pubmed.ncbi.nlm.nih.gov/42317859/

Overview of Files:

Standalone Files:

• bl_utils.py: Assorted helper functions and objects
• cleaning.ipynb: Initial data cleaning, formatting, and train/test splitting
• preprocessing.ipynb: Downstream data scaling, collinearity removal, & feature selection
• ml_baseline.ipynb: Machine learning benchmarking for the dataset
• modeling_BLIMP_exp: Modeling of BLIMP-E (expression data only)
• modeling_BLIMP_exp: Modeling of BLIMP-M (mutation data only)
• modeling_BLIMP_full.ipynb: Modeling of BLIMP with all modalities, SHAP analysis, and final results

“data” Folder:

• mbn_sfu_2023: Raw data directly downloaded from CBioPortal
• cleaning_output: Output files from cleaning.ipynb notebook
• preprocessing_output: Output files from preprocessing.ipynb
• ml_output: Output files from ml_baseline.ipynb notebook
• modeling_output: Output files from modeling_BLIMP_full.ipynb, modeling_BLIMP_exp.ipynb and modeling_BLIMP_mut.ipynb notebooks
• burkitt_genes.txt: gene HUGO nomenclature names of genes known to be involved in BL pathology (gathered from literature review)
• clinica_variant_annotationl.tsv: Annotations of patients sporadic/endemic classification
• gene_lengths.txt: Gene lengths downloaded from http://useast.ensembl.org/biomart
• HGNC.txt: HUGO nomenclature information downloaded from http://useast.ensembl.org/biomart