Machine Learning Models Utilizing Oxidative Stress Biomarkers for Breast Cancer Prediction: Efficacy and Limitations...

Objective: This study aimed to apply the Random Forest machine learning model using oxidative stress biomarkers to classify breast cancer status and assess sentinel lymph node (SLN) metastasis, a pathology of high incidence and mortality that represents a major public health challenge. Methods: The breast cancer classification cohort included 188 women with infiltrating ductal carcinoma and 78 healthy volunteers. For SLN metastasis assessment, a subset of 29 women with metastases and 57 controls (n = 86) was used. Data preprocessing and the SMOTE technique were applied to balance the classes in the metastasis set, achieving a perfect balance of 171 examples (57 per class). Random Forest model with a leave-one-out validation strategy was employed and oxidative stress biomarkers (e.g., lipid peroxidation, total antioxidant capacity, superoxide dismutase, catalase, glutathione peroxidase) were used. Results: The model achieved high accuracy (0.996) in classifying breast cancer, representing a substantial improvement over current screening methods such as mammography. In contrast, its performance in detecting SLN metastases was more limited (accuracy = 0.854), likely reflecting the inherent complexity and heterogeneity of the metastatic process. Moreover, these estimates derive from a retrospective case-control cohort and should not be viewed as a substitute for, or a direct comparison with, population-based mammography screening, which would require dedicated prospective validation. Conclusions: The findings underscore the model's robust performance in distinguishing women with breast cancer from healthy volunteers, but highlight significant gaps in its ability to diagnose metastatic disease. Future research should integrate additional biomarkers, longitudinal data, and explainable artificial intelligence (XAI) methods to improve clinical interpretability and accuracy in metastasis prediction, moving towards precision medicine.