Breast cancer prediction based on gene expression data using interpretable machine learning techniques

Area of Science:

• Oncology
• Bioinformatics
• Computational Biology

Background:

• Breast cancer is a leading cause of cancer deaths globally.
• Accurate prediction and diagnosis are crucial for effective treatment and patient survival.
• Machine learning (ML) offers potential for improving breast cancer prediction accuracy.

Purpose of the Study:

• To accurately predict breast cancer using ML models.
• To identify influential predictive genes through feature selection.
• To enhance model interpretability using explainable ML techniques.

Main Methods:

• Utilized a dataset of 1208 observations and 3602 genes.
• Employed feature selection techniques and ML models: K-nearest Neighbors (KNN), Random Forests (RF), and Support Vector Machine (SVM).
• Applied explainable ML methods (Shapley values, PDPS, ALE plots) and model-based ranking (LOCI) for gene importance analysis.

Main Results:

• Identified key genes crucial for breast cancer prediction using Shapley values and the LOCI method.
• Achieved aligned gene rankings from SVM and RF models via LOCI.
• Visualizations (PDPS, ALE plots) illustrated feature effects and interactions, confirming model interpretability.

Conclusions:

• Machine learning models, combined with feature selection and explainable AI, provide interpretable and reliable breast cancer prediction.
• Explainable ML approaches are vital for medical decision-making in oncology.
• This study highlights the potential of integrating advanced computational methods for improved breast cancer diagnostics.