A Predictive Model for Personalized Therapeutic Interventions in Non-small Cell Lung Cancer

Area of Science:

• Oncology
• Bioinformatics
• Computational Biology

Background:

• Non-small cell lung cancer (NSCLC) is the most common lung cancer, often diagnosed at advanced stages.
• Molecular targeted therapies improve survival and quality of life, but personalized treatment faces challenges in data integration and clinical decision support.
• Effective patient selection for targeted therapy is essential for optimizing treatment outcomes in advanced NSCLC.

Purpose of the Study:

• To identify relationships between patient characteristics and tumor response in advanced NSCLC using frequent pattern mining.
• To develop and evaluate machine learning classifiers for predicting treatment outcomes in patients receiving EGFR tyrosine kinase inhibitors.
• To explore the potential of computational approaches for clinical decision support in personalized NSCLC therapy.

Main Methods:

• Frequent pattern mining was employed to analyze associations between patient data and treatment response.
• Univariate analysis identified significant factors including smoking status, histology, EGFR mutation, and targeted drug.
• Four machine learning classifiers were applied to predict treatment outcomes from EGFR tyrosine kinase inhibitor therapy.

Main Results:

• Univariate analysis revealed significant associations between smoking status, histology, EGFR mutation, and response to targeted therapy.
• The highest classification accuracy achieved was 76.56% with an area under the curve of 0.76.
• A decision tree model effectively predicted tumor response using EGFR mutations, histology, and smoking status, providing interpretable results.

Conclusions:

• Machine learning models, particularly decision trees and support vector machines, show promise for clinical decision support in advanced NSCLC.
• Integrating clinical and genetic data through computational methods can aid in selecting appropriate patients for targeted therapies.
• These findings support the development of advanced decision support tools for personalized treatment selection in NSCLC.