Metabolomic Profiling and Machine Learning-Based Prediction of High-Dose Methotrexate-Induced Liver Injury in
Kai Guo1,2, Lei Wang3, Xiaoran Feng1,2
1National Center for Clinical Laboratories, Institute of Geriatric Medicine, Chinese Academy of Medical Sciences, Beijing Hospital/National Center of Gerontology, Beijing, China.
Pediatric Blood & Cancer
|July 23, 2026
Summary
Metabolomic profiling can predict high-dose methotrexate (HD-MTX) liver injury in pediatric acute lymphoblastic leukemia (ALL) patients. Machine learning models using these biomarkers achieve high accuracy for early risk assessment and personalized treatment strategies.
Area of Science:
- Biochemistry
- Oncology
- Computational Biology
Background:
- High-dose methotrexate (HD-MTX) is a critical treatment for pediatric acute lymphoblastic leukemia (ALL).
- HD-MTX commonly causes hepatotoxicity, necessitating early prediction tools.
- This study addresses the need for early detection of HD-MTX-induced liver injury.
Purpose of the Study:
- To identify predictive metabolomic biomarkers for HD-MTX-induced hepatotoxicity in pediatric ALL.
- To develop machine learning models for early clinical risk assessment.
- To optimize therapeutic strategies through precise risk stratification.
Main Methods:
- 106 pediatric ALL patients undergoing HD-MTX therapy were studied.
- Untargeted plasma metabolomics was performed pre- and post-treatment.
- Machine learning algorithms (Random Forest, SVM, Bayesian logistic regression) were developed for risk prediction, with SHAP analysis for interpretability.
Main Results:
- Distinct metabolic signatures differentiated patients with and without liver injury.
- Pre-treatment metabolites, including glycocholic acid, predicted liver injury (altered arginine biosynthesis, glutathione metabolism).
- Post-treatment metabolites, including arachidic acid, further distinguished the injury group; ML models achieved AUC > 0.900.
Conclusions:
- Metabolic alterations characterize HD-MTX-induced liver injury in pediatric ALL.
- A metabolomics-based framework enables early risk assessment for hepatotoxicity.
- These findings support precision therapeutic strategies for pediatric ALL patients.
