Use of Machine Learning and Multi-Omics to Discover Ferroptosis-Driven Mechanisms for Breast Cancer Prognosis

Use of Machine Learning and Multi-Omics to Discover Ferroptosis-Driven Mechanisms for Breast Cancer Prognosis

Breast cancer remains a leading cause of cancer-related deaths among female patients globally. Ferroptosis, which is a recently identified form of regulated cell death, is characterized by the accumulation of iron-dependent lipid peroxidation and plays a pivotal role in various pathophysiological conditions, including breast cancer. Machine learning algorithms are used to identify prognostic biomarkers for breast cancer patients by processing complex, high-dimensional biological data and discovering meaningful patterns.
However, there is little evidence that machine learning methods identify critical ferroptosis-associated genes for breast cancer prognosis. For this study, clinical and genetic data are based on The Cancer Genome Atlas (TCGA). Differential gene expression is assessed using the GSCA platform. Important genes are selected using an Elastic Net algorithm. A prognostic gene panel is established using the Kaplan-Meier (K-M) Plotter. Additional analyses, including a Cox proportional hazards model, differential gene expression heatmaps grouped by specific genes and tumor immune microenvironment (TIME) evaluation using the TCGAplot R package.
The biological relevance of these differentially expressed genes is determined using Gene Set Enrichment Analysis (GSEA). Pharmacogenetic analysis uses Q-omics. Seven differentially expressed ferroptosis-associated genes are identified: SLC7A11, FANCD2, CISD2, VDAC1, VDAC3, CISD1 and GPX4. Elastic Net identifies four significant prognostic genes: CISD1, VDAC1, VDAC2 and VDAC3. A risk score model that uses these four genes outperforms individual genes in terms of prognostication and CISD1 demonstrates the highest weight coefficient. CISD1 has the most significant impact on the hazard ratio for the breast cancer cohort, compared to other cancer types. GSEA and mutation landscape profiling respectively show that CISD1-associated gene networks are involved in cell cycle regulation and driver gene mutations.
TIME analysis shows that there is a positive correlation between CISD1 expression and immune checkpoint-related genes, which is indicative of an immunosuppressive microenvironment. Breast cancer cell lines with elevated CISD1 levels exhibit resistance to cetuximab and demonstrate sensitivity to sorafenib.
This study identifies a prognostic panel using four ferroptosis-associated genes. CISD1 contributes the highest prognostic weight. Elevated CISD1 levels are associated with poor outcomes, activation of the cell cycle, mutations in driver genes and an immunosuppressive microenvironment. CISD1 expression also acts as a biomarker for the development of precision medicine and can be used to guide therapeutic interventions for breast cancer.