Refining risk prediction in pediatric acute lymphoblastic leukemia through DNA methylation profiling

Area of Science:

• Oncology
• Epigenetics
• Bioinformatics

Background:

• Acute lymphoblastic leukemia (ALL) is the most common childhood cancer.
• Relapses and mortality remain significant challenges despite treatment advances.
• Predictive tools are needed to refine risk stratification and personalize therapy.

Purpose of the Study:

• To develop and validate machine learning models for predicting relapse and mortality risk in pediatric ALL using DNA methylation data.
• To assess the prognostic value of DNA methylation in independent patient cohorts.
• To explore the integration of epigenetic predictors with clinical factors for improved risk assessment.

Main Methods:

• Supervised machine learning (random survival forests) applied to array-based DNA methylation data.
• Development of a relapse risk predictor (RRP) using 16 CpG sites and a mortality risk predictor (MRP) using 53 CpG sites.
• Validation in independent Nordic and Canadian pediatric ALL cohorts.

Main Results:

• The RRP and MRP demonstrated good predictive performance in training and test sets (c-indexes ranging from 0.667 to 0.754).
• External validation confirmed the prognostic value of RRP and MRP in independent cohorts.
• Integration with traditional risk groups improved model precision; MRP identified a high-risk subgroup.

Conclusions:

• DNA methylation serves as a valuable prognostic factor in pediatric ALL.
• Epigenetic profiling can refine risk stratification, potentially leading to personalized treatment strategies.
• Machine learning models integrating DNA methylation data offer a promising approach for managing pediatric ALL.