Akademik Veri Yönetim Sistemi
16.Tıp Bilişimi Kongresi, Ankara, Türkiye, 22 - 23 Mayıs 2025, ss.55, (Özet Bildiri)
Accurate prediction of metastasis in thyroid cancer is critical for de- termining prognosis and developing effective treatment strategies. In this study, we aimed to predict metastasis status (M1 vs. M0) using clinical data, miRNA expression profiles, and BRAF mutation status. Mutation data, miRNA-seq, and clinical information were obtained and integrated from The Cancer Genome At- las (TCGA) database. Differential expression analysis was performed to identify miRNAs significantly associated with metastatic and non-metastatic groups. Ma- chine learning models, including Random Forest, XGBoost, and Decision Tree, were trained using the selected features. Among the models, XGBoost demon- strated the best performance with an ROC AUC of 0.993, a recall of 0.98, and an F1-score of 0.96. These findings suggest that miRNAs have strong potential as biomarkers for metastasis and that predictive models developed through inte- grated omics analysis can be utilized in clinical decision support systems. Early prediction of the risk of metastasis will provide a critical advantage for early di- agnosis and timely intervention by allowing close follow-up of patients.