Pathway-Informed Machine Learning Identifies Genetic Predictors of High-Dose Methotrexate-Induced Mucositis in...

Area of Science:

• Oncology
• Pharmacogenomics
• Genetics

Background:

• High-dose methotrexate is crucial for pediatric cancer treatment, but frequently causes mucositis.
• Mucositis can lead to treatment delays, dose reductions, and negatively impact patient survival.
• The cumulative genetic influence on methotrexate-induced mucositis across biological pathways is not well understood.

Purpose of the Study:

• To investigate the impact of genetic variations within biological pathways on methotrexate-induced mucositis in pediatric cancer patients.
• To develop a predictive model for mucositis risk based on genetic polymorphisms.

Main Methods:

• Evaluated single nucleotide polymorphisms (SNPs) across 18 relevant pathways in 278 pediatric acute lymphoblastic leukemia patients.
• Utilized the Joint Association of Genetic variants tool for pathway enrichment analysis.
• Developed a predictive model using XGBoost machine learning to assess SNP contributions.

Main Results:

• Identified significant pathway enrichment in IL6 (P=0.04) and WNT/β-catenin (P=0.048) signaling.
• The predictive model achieved an area under the curve (AUC) of 0.76, highlighting SNPs in inflammation and mucosa-related genes (e.g., PRKCD, IL17B).
• Model performance significantly decreased (AUC dropped to 0.61) when SNP features were removed, confirming their predictive value.

Conclusions:

• This pathway-informed genetic analysis identifies key contributors to methotrexate-induced mucositis.
• The findings support the use of polygenic risk prediction for individualized toxicity profiling in pediatric oncology.
• This approach may guide the development of targeted therapies to mitigate treatment-limiting mucositis.