Integrating single-cell RNA and bulk RNA sequencing data to identify prognostic genes associated with pyrimidine metabolism in triple-negative breast cancer by machine learning algorithm combinations

BACKGROUND: Pyrimidine metabolism plays a crucial role in DNA synthesis and cell proliferation and is associated with the development of various cancers. However, their prognostic value in triple-negative breast cancer remains to be further investigated. This study aimed to identify the prognostic genes related to pyrimidine metabolism in triple-negative breast cancer (TNBC). METHODS: All data applied in this study were obtained from public databases. The prognostic genes related to pyrimidine metabolism in TNBC were identified through Weighted gene co-expression network analysis (WGCNA), differential expression analysis, univariate Cox regression, and 101 combinations of machine learning algorithms. Based on the optimal model among the 101 combinations of machine learning algorithms, a risk model was built. Then, the risk model was validated in GSE58812 dataset. Subsequently, a nomogram was built based on prognostic genes. Meanwhile, enrichment analysis, gene mutation analysis, and drug sensitivity analysis were also executed. In addition, the somatic mutation analysis of TNBC patients revealed that missense mutations were the predominant mutation type. To explore the cells that influence the risk of TNBC, single-cell RNA sequencing analysis was conducted. Finally, the expression levels of prognostic genes were verified through qPCR experiments. SEPT3 expression in TNBC cells was detected by qPCR; its effects on TNBC cell proliferation (CCK-8) and invasion/migration (Transwell, wound-healing) were evaluated. RESULTS: Firstly, 5 prognostic genes (ECE2, NFE2L3, PFKFB3, FADS2, and SEPT3) were identified. The risk model demonstrated high reliability, and the survival rate of high-risk patients was relatively low. A nomogram with relatively high prognostic accuracy was constructed. Next, 22 pathways were identified by enrichment analysis. A drug sensitivity analysis identified 12 drugs were identified (e.g. ABT.263). In addition, epithelial cells were related to the expression of the prognostic genes. CONCLUSION: ECE2, NFE2L3, PFKFB3, FADS2, and SEPT3 are associated with pyrimidine metabolism in TNBC. A risk model and nomogram were successfully constructed based on these genes, providing a theoretical basis for the treatment of TNBC patients. We confirm the model's validity in TNBC by validating SEPT3 (a key PyMRG) regulates TNBC cell proliferation, migration, and invasion.