Integrated Bioinformatics and Machine Learning Analysis Identifies MT1F as a Potential Diagnostic Biomarker and Therapeutic Target in Breast Cancer

Abstract

Breast cancer (BRCA) is a heterogeneous tumour and is the most common cancer in the world, and there is a need for strong biomarkers to enhance diagnosis and targeted therapy. We created a combined bioinformatics and machine learning approach for identifying important molecular biomarkers related to BRCA in this study. Two gene expression datasets were analyzed and 2,386 commonly dysregulated genes were identified. Enrichment analysis of functions indicated that extracellular matrix organization and immune related pathways were significantly involved. Protein protein interaction (PPI) network analysis revealed that MT1F and CCNA2 were hub genes in the network, and MT1F was consistently ranked among the most central genes by various topological and machine learning techniques. The machine learning models (Random Forest, Gradient Boosting, and XGBoost) showed high diagnostic performance, with the Random Forest model having the highest discriminative ability (AUC = 0.990). It was found that MT1F was significantly upregulated in many different malignancies by pan-cancer analysis and was epigenetically regulated by DNA methylation analysis. Immune infiltration analysis also showed significant correlations between the expression of MT1F and immune cell populations. The drug sensitivity analysis (DSA) with GSCA datasets showed that MT1F expression was significantly correlated with sensitivity to several anti cancer drugs, suggesting its role as a potential predictive biomarker. To conclude, MT1F might be a diagnostic biomarker and therapeutic target for breast cancer. This study highlights the value of integrative computational methods in the discovery of biomarkers and precision oncology.