Machine learning approaches for predicting progression in hormone-sensitive prostate cancer patients.

Area of Science:

• Oncology
• Bioinformatics
• Machine Learning

Background:

• Hormone-sensitive prostate cancer (HSPC) frequently progresses to castration-resistant prostate cancer (CRPC) after androgen deprivation therapy (ADT).
• Predicting this progression is crucial for timely clinical intervention and treatment planning.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for predicting HSPC progression to CRPC.
• To identify significant clinical features and markers associated with this transition using statistical analysis.

Main Methods:

• Analysis of clinical data from 410 HSPC patients.
• Application of ML models including decision tree, random forest, XGBoost, ANN, and SVM.
• Feature selection using a genetic algorithm (GA) and model evaluation via AUC, calibration plots, and learning curves.

Main Results:

• Ensemble learning methods, specifically Random Forest (RF) and XGBoost, demonstrated superior performance.
• RF achieved an AUC of 0.873 on the test set, while XGBoost achieved an AUC of 0.866.
• Models showed good calibration and no significant overfitting, indicating reliable predictive capabilities.

Conclusions:

• Ensemble ML methods, particularly Random Forest, are effective in predicting HSPC progression.
• Baseline clinical data holds potential for risk stratification of HSPC patients.
• Further validation in larger, multi-center prospective studies is recommended for clinical integration.