Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Predicting Survivability of Cancer Patients with Metastatic Patterns Using Explainable AI

Polycarp Nalela, Deepthi Rao, Praveen Rao

TL;DR

This study addresses prognostic prediction for cancer patients with metastatic patterns by leveraging the MSK-MET pan-cancer dataset. It compares five ML models, finds XGBoost to be the best performer with an AUC of $0.82$, and uses SHAP to reveal interpretable predictors such as metastatic site count, tumor mutation burden, and fraction of genome altered. The authors extend the analysis with survival modeling (Kaplan-Meier, Cox PH, and XGBoost Survival Analysis) to connect predictions with time-to-event outcomes, demonstrating clinically actionable insights and cancer-type–specific differences. The work emphasizes explainability and the potential for personalized prognosis and treatment planning, highlighting both global patterns and cancer-specific nuances. Overall, the integrated predictive and survival framework offers a robust, interpretable tool to improve patient care in metastatic cancer settings.

Abstract

Cancer remains a leading global health challenge and a major cause of mortality. This study leverages machine learning (ML) to predict the survivability of cancer patients with metastatic patterns using the comprehensive MSK-MET dataset, which includes genomic and clinical data from 25,775 patients across 27 cancer types. We evaluated five ML models-XGBoost, Naïve Bayes, Decision Tree, Logistic Regression, and Random Fores using hyperparameter tuning and grid search. XGBoost emerged as the best performer with an area under the curve (AUC) of 0.82. To enhance model interpretability, SHapley Additive exPlanations (SHAP) were applied, revealing key predictors such as metastatic site count, tumor mutation burden, fraction of genome altered, and organ-specific metastases. Further survival analysis using Kaplan-Meier curves, Cox Proportional Hazards models, and XGBoost Survival Analysis identified significant predictors of patient outcomes, offering actionable insights for clinicians. These findings could aid in personalized prognosis and treatment planning, ultimately improving patient care.

Predicting Survivability of Cancer Patients with Metastatic Patterns Using Explainable AI

TL;DR

This study addresses prognostic prediction for cancer patients with metastatic patterns by leveraging the MSK-MET pan-cancer dataset. It compares five ML models, finds XGBoost to be the best performer with an AUC of , and uses SHAP to reveal interpretable predictors such as metastatic site count, tumor mutation burden, and fraction of genome altered. The authors extend the analysis with survival modeling (Kaplan-Meier, Cox PH, and XGBoost Survival Analysis) to connect predictions with time-to-event outcomes, demonstrating clinically actionable insights and cancer-type–specific differences. The work emphasizes explainability and the potential for personalized prognosis and treatment planning, highlighting both global patterns and cancer-specific nuances. Overall, the integrated predictive and survival framework offers a robust, interpretable tool to improve patient care in metastatic cancer settings.

Abstract

Cancer remains a leading global health challenge and a major cause of mortality. This study leverages machine learning (ML) to predict the survivability of cancer patients with metastatic patterns using the comprehensive MSK-MET dataset, which includes genomic and clinical data from 25,775 patients across 27 cancer types. We evaluated five ML models-XGBoost, Naïve Bayes, Decision Tree, Logistic Regression, and Random Fores using hyperparameter tuning and grid search. XGBoost emerged as the best performer with an area under the curve (AUC) of 0.82. To enhance model interpretability, SHapley Additive exPlanations (SHAP) were applied, revealing key predictors such as metastatic site count, tumor mutation burden, fraction of genome altered, and organ-specific metastases. Further survival analysis using Kaplan-Meier curves, Cox Proportional Hazards models, and XGBoost Survival Analysis identified significant predictors of patient outcomes, offering actionable insights for clinicians. These findings could aid in personalized prognosis and treatment planning, ultimately improving patient care.

Paper Structure

This paper contains 29 sections, 10 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Data pre-processing
  4. Stratified random sampling of training and testing sets
  5. Selection of machine learning models
  6. Grid-search with hyperparameter tuning
  7. Model architecture
  8. Evaluation of the models
  9. Model interpretation and explanation
  10. Survival analysis
  11. Kaplan-Meier Survival Analysis
  12. Cox Proportional Hazards Model
  13. Log-Rank Test
  14. XGBoost Survival Analysis
  15. Results
  16. ...and 14 more sections

Figures (10)

  • Figure 1: An abstract implementation of the ML models for predicting metastatic cancer survivability
  • Figure 2: AUC comparison for the five ML models in the ensemble setup
  • Figure 3: SHAP-Beeswarm plot for Unified XGBoost model
  • Figure 4: SHAP-Beeswarm plot for Non Small Cell Lung Cancer
  • Figure 5: SHAP-Beeswarm plot for Colorectal Cancer
  • ...and 5 more figures