Identification of copper related biomarkers in breast cancer using machine learning

Abstract

Background: Breast cancer is the most prevalent and deadly cancer among women globally, necessitating more effective diagnostic and therapeutic approaches. This study aims to explore new treatment targets and diagnostic tools.

Methods: Employing machine learning techniques and utilizing PCR, IHC technologies, and multiple databases, we identified and validated genes closely linked with breast cancer and copper-induced cell death. We then explored how their expression levels impact cancer diagnosis, prognosis, immune cell infiltration, and drug sensitivity.

Results: This investigation identified three crucial genes-MT1M, GRHL2, and PKM-intimately associated with the copper death mechanism in breast cancer pathology. Validated through comprehensive analysis across cells, tissue models, and diverse databases, these genes showed significant differential expression (P-value < 0.05), affirming their pivotal role in enhancing diagnostic accuracy (AUC values: 0.917, 0.970, 0.951) and prognostic assessment (HR = 0.65, P = 0.018; HR = 1.69, P = 0.0011; HR = 1.51, P = 0.012) in breast cancer. Additionally, their expression levels influence the infiltration of immune cells and the sensitivity to certain drugs.

Conclusion: MT1M, GRHL2, and PKM are novel diagnostic and therapeutic targets for breast cancer. These findings enhance prognostic evaluations, deepen our understanding of its mechanisms.

Keywords: Biomarker; Breast cancer; Copper death; Immune infiltration; Machine learning.

Fig. 1

Flowchart for research

Fig. 2

Analysis of Differentially Expressed Genes (DEGs) and Immune Infiltration. A The volcano plot presents DEGs in dataset GSE42568. B A heatmap displays the pattern of these DEGs. C Venn diagrams illustrate the overlap of DEGs with genes related to Copper death (DEGCUs). D Immune infiltration analysis explores the correlation between 22 types of immune infiltrating lymphocytes and DEGCUs in GSE42568, comparing control (C) and patient (P) samples.

Fig. 3

Construction of weighted gene co-expression networks. A Sample clustering dendrogram with tree leaves corresponding to individual samples. B Shows the original and combined modules under the clustering tree. C Heat map of module-trait relationships, where each color represents a co-expression module and the values represent module-trait correlation coefficients and p-values. It can be seen that the ME turquoise module has the highest correlation with breast cancer. WGCNA, weighted gene co-expression networks analysis

Fig. 4

Acquisition and functional enrichment analysis of 79 DEGCUs. A Venn diagrams of 79 DEGCUs. B, C The results of GO and KEGG analysis are displayed by bubble plots, respectively. GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; BP, biological process; CC, cellular component; MF, molecular function

Fig. 5

Screening hub genes by machine learning. A LASSO regression algorithm. B SVM-RFE algorithm. C RF algorithm. D Venn diagrams for three algorithms. LASSO, Least Absolute Shrinkage and Selection Operator; SVM-RFE, Support Vector Machine-Recursive Feature Elimination; RF, Random Forest

Fig. 6

ROC Analysis, GSEA, and PCR Validation. A ROC curve analysis. Assesses the predictive values of these genes using ROC curves from the GSE42568 dataset. B GSEA analysis of hub genes. C Real-Time PCR. Quantifies mRNA levels of the genes in breast cancer cells versus normal cells, with statistical significance marked as *P-value < 0.05, **P-value < 0.01, ***P-value < 0.001;'ns' for no significant difference

Fig. 7

Immunohistochemical Staining Results and Statistical Analysis of Key Genes in Infiltrating Breast Cancer. A Immunohistochemical images and local magnification of infiltrating breast cancer and adjacent non-cancerous tissues. B Statistical results of gene expression in infiltrating breast cancer and adjacent non-cancerous tissues