Discovering new genetic markers for breast cancer diagnosis via advanced machine learning techniques

---

Presentation Time: 03:15 PM - 04:30 PM
Poster Board Number: B906
Abstract ID: 5197

Presenting Author:
Lingzhen Wang , Student at Western Reserve Acad.

Abstract:

Objectives: This study had two goals: 1) To assess 15 machine learning models using a comprehensive dataset to identify the most effective one based on performance metrics; and 2) To use the insights from this model to discover new genetic markers for breast cancer diagnosis.

Methods: We evaluated 15 machine learning models, focusing on accuracy, precision, and Area Under the Receiver Operating Characteristic Curve (AUC). The top model was then used for advanced analysis to explore potential breast cancer genetic markers.

Results: One machine learning model excelled in accuracy and AUC, effectively differentiating between normal and cancerous gene expressions. SHapley Additive exPlanations (SHAP) analysis identified a set of genes as potential biomarkers, correlating with known breast cancer genetic factors and demonstrating the method's effectiveness in finding relevant markers.

Conclusion: The study leveraged machine learning to uncover new genetic biomarkers for breast cancer, with the selected model revealing key genes through SHAP analysis. These findings highlight the utility of machine learning in medical diagnostics and pave the way for better breast cancer detection and treatment.