Abstract 14410: A Machine Learning Approach to Predicting Risk of Anthracycline Cardiotoxicity in Pediatric Cancer Survivors

Background: Despite known clinical and genetic risk factors, predicting anthracycline cardiotoxicity remains challenging. Objective: To develop a risk prediction model for anthracycline cardiotoxicity in childhood cancer survivors. Methods: We performed exome sequencing in 289 childhood cancer survivors at least 3 years from anthracycline exposure. In a nested case-control design, 183 cases with LV ejection fraction (LVEF) ≤55% despite low-dose doxorubicin (DOX) (≤250 mg/m2), and 106 controls with LVEF >55% despite DOX >250 mg/m2 were selected as extreme phenotypes. Rare/low-frequency variants were collapsed to identify genes differentially enriched for variants between cases and controls. The top-ranked genes were functionally evaluated in human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs) and variant enrichment was confirmed in a replication cohort. Using random forest, a risk prediction model that included genetic and clinical predictors was developed. Results: Thirty-one genes were differentially enriched for variants between cases and controls (p<0.001). Only 42.6% cases harbored a variant in these genes compared to 89.6% controls [Odds ratio (95% CI), 0.086 (0.043-0.171), p=3.98x10 -15 ] suggesting that absence of these variants may predispose to cardiotoxicity. A risk prediction model for cardiotoxicity that included clinical and genetic factors had a higher prediction accuracy and lower misclassification rate compared to the clinical model. In vitro inhibition of gene-associated pathways ( PI3KR2, ZNF827 ) provided protection from cardiotoxicity in CMs. Conclusions: Our study identified variants that protect against cardiotoxicity and informed the development of a prediction model for delayed anthracycline cardiotoxicity, and also provided new targets in autophagy genes for development of cardio-protective drugs.