Machine Learning-Driven Integration of Cancer Cell Phenotypes Predicts Cisplatin Sensitivity.

Area of Science:

• Oncology
• Bioinformatics
• Genomics

Background:

• Precision medicine has revolutionized cancer therapy, primarily through genomic profiling for targeted drugs and immunotherapies.
• Current genomic tests are limited in predicting efficacy for traditional anticancer agents.
• A novel phenotype-based classification method using gene expression is proposed to predict classical anticancer agent effectiveness.

Purpose of the Study:

• To develop a machine learning model for predicting sensitivity to classical anticancer agents based on gene expression patterns.
• To establish a phenotype-based classification system for personalized chemotherapy selection.

Main Methods:

• Hierarchical clustering of IC50 values to differentiate cisplatin-sensitive and resistant cell lines.
• Differentially expressed gene (DEG) analysis combined with SHAP value-based machine learning to identify key predictive genes.
• Development of the Cisplatin Sensitivity Predictor using 26 Genes (CSP26G) model.

Main Results:

• The CSP26G model demonstrated external validity in cisplatin-resistant cell lines (A549CR).
• The model successfully classified non-small cell lung cancer patients from TCGA into sensitive and resistant groups, correlating with survival outcomes.
• CSP26G showed predictive capability for cisplatin and other DNA-damaging agents.

Conclusions:

• Integrating DEG analysis and machine learning enables a robust drug sensitivity prediction model.
• This model advances personalized precision medicine for classical chemotherapies.
• The findings support the clinical application of gene expression-based prediction for optimizing chemotherapy regimens.