A public, high-level view of the approach
This page summarizes the analytical approach at a non-sensitive level. It omits internal implementation detail, data-access instructions, and unreleased code.
Four stages, one reproducible workflow
The project works with publicly accessible Burkitt lymphoma datasets, analyzed in R and Python using version-controlled, containerized workflows so results can be reproduced from documented code.
Data
Publicly accessible genomic, transcriptomic, and clinical datasets, harmonized to a common schema.
Pathway scoring
Sample-level metabolic pathway activity inferred from gene expression with GSVA.
Feature engineering
Composite metabolic indices and endemic-context indicators built from the literature.
Modeling
Multivariate survival modeling, validated and benchmarked against the clinical baseline.
Turning expression into interpretable pathway scores
Metabolic pathway activity is inferred from gene-expression data using Gene Set Variation Analysis (GSVA) against curated pathway gene sets, including KEGG, Reactome, and custom Burkitt-lymphoma-relevant collections.
Differences across genetic subgroups are assessed with linear models and multiple-testing correction, and analyses are stratified by EBV status to separate subgroup-associated signals from EBV-associated ones.
Composite indices grounded in oncogenic biology
Literature-derived composite metabolic indices summarize biology linked to the disease's principal oncogenic drivers — MYC activity, EBV-associated reprogramming, B-cell-receptor and downstream signaling, and tumor-suppressor-related disruption.
Endemic-context indicators represent malaria-related and geographic factors where appropriate metadata are available, with explicit handling of missing values.
Survival modeling, benchmarked honestly
Risk modeling is planned using multivariate survival methods, including regularized Cox regression with survival-forest and gradient-boosting comparators. Performance is evaluated with nested cross-validation, assessed for discrimination and calibration, and benchmarked against a clinical-variable-only baseline.
Methods described here as planned should be read as the intended approach, not completed findings.
Built to be checked and reused
Version control
Code and analysis notebooks tracked so every result traces back to its source.
Containerization
Portable environments so analyses run the same way across machines.
Curated gene sets
Documented pathway gene sets with literature citations.
Open source
Approved, public-facing materials released openly for review.