Prediction of Breast Cancer Distant Metastasis by Artificial Intelligence Methods from an Epidemiological Perspective

Abstract

Introduction: Despite significant advances in breast cancer (BC) management, the prognosis for most patients with distant metastasis remains poor. We predicted distant metastasis in BC patients with artificial intelligence (AI) methods based on genomic biomarkers. Methods: The dataset used in the study included 97 patients with BC, of whom 46 (47%) developed distant metastases, and 51 (53%) did not develop distant metastases, and the expression level of 24,481 genes of these patients. An approach combining Boruta + LASSO methods was applied to identify biomarker genes associated with BC distant metastasis. Mann-Whitney U test was used to examine the difference between groups in terms of gene expression levels in statistical analyses, and Cohen d effect sizes and odds ratios were calculated. AdaBoost and XGBoost algorithms, which are tree-based methods, were used for BC distant metastasis prediction, and the results were compared by evaluating comprehensive performance criteria. Results: After Boruta + LASSO methods, 14 biomarker candidate genes were identified. These predictive genes were PIB5PA, SSX2, OR1F1, ALDH4A1, FGF18, WISP1, PRAME, CEGP1, AL080059, NMU, ATP5E, SMARCE1, FGD6, and SLC37A1 . In effect size results; in particular, show that the AL080059 (Cohen’s D: 1.318) gene is clinically predictive of BC Metastasis. The accuracy, F1-score, positive predictive value, sensitivity, and area under the ROC Curve (AUC) values obtained with the AdaBoost algorithm for BC metastasis prediction was 95%, 96.3%, 100%, 92.6%, and 98.8%, respectively. The model created with the XGBoost algorithm, on the other hand, obtained 90%, 92.9%, 92.9%, 92.9%, 97.6% accuracy, F1-score, positive predictive value, sensitivity, and AUC values, respectively. Conclusion: Identifying genes that successfully predict BC distant metastasis with AI methods in the study may be decisive for future therapeutic targets and help clinicians better adapt adjuvant chemotherapy to their patients. Additionally, the AdaBoost prediction model created can discriminate patients at risk of BC distant metastases.