Machine Learning Identifies Clinical and Genetic Factors Associated With Anthracycline Cardiotoxicity in...

Area of Science:

• Genomics
• Cardiology
• Pediatric Oncology

Background:

• Predicting anthracycline cardiotoxicity, a serious side effect of cancer treatment, remains difficult despite known clinical risk factors.
• Childhood cancer survivors are at long-term risk for cardiac complications following anthracycline chemotherapy.

Purpose of the Study:

• To develop a robust clinical and genetic risk prediction model for anthracycline cardiotoxicity in pediatric cancer survivors.
• To identify genetic variants associated with differential susceptibility to anthracycline-induced heart damage.

Main Methods:

• Exome sequencing was performed on 289 childhood cancer survivors with varying doxorubicin exposure and cardiac outcomes.
• A nested case-control design identified extreme phenotypes of cardiotoxicity and preserved cardiac function.
• Random forest modeling integrated clinical and genetic predictors to create a risk prediction model.

Main Results:

• Thirty-one genes showed differential variant enrichment between cases and controls (p < 0.001).
• A combined clinical and genetic model demonstrated superior prediction accuracy and reduced misclassification compared to a clinical-only model.
• In vitro studies showed that inhibiting PI3KR2 and ZNF827 pathways protected cardiomyocytes from cardiotoxicity.

Conclusions:

• Genetic variants in cardiac injury pathway genes are associated with protection against anthracycline cardiotoxicity.
• The developed prediction model improves the identification of survivors at risk for delayed cardiotoxicity.
• Autophagy-related genes present potential therapeutic targets for developing cardio-protective drugs.