Single-cell sequencing and transcriptome analysis identified a comprehensive risk score model based on immune-related metabolic genes for improving the prediction of sensitivity in breast cancer immunotherapy

Background: The metabolic reprogramming of breast cancer (BC) has gained great attention in recent years. Malignant and infiltrating immune cells compete for nutrients and metabolites; still, the impact of metabolism on them remains to be further elucidated. The specific objective of this analysis was to anatomy the action of immune-related metabolic genes in breast cancer and develop a combined model to predict susceptibility to immunotherapy, thus helping guide patient management and establish personalized risk assessment with superior accuracy and clinical applicability. Methods: This study was based on data of 1048 BC patients from The Cancer Genome Atlas (TCGA). 46 immune-related metabolic genes were identified by differential expression analysis between different tissue states. Applying unsupervised clustering and other bioinformatics techniques, we illustrated how the divergent groups' immunometabolism and survival conditions varied. A comprehensive risk-sharing index model was developed using LASSO regression and multivariable Cox analysis method, and BC patients were categorized into two risk groups based on their levels of risk score. Another three independent GEO database sets [GSE20685, GSE42568, GSE124647] were selected for external validation. Finally, the single-cell sequencing data mining and analysis aimed to explore the immunometabolic heterogeneity of human breast cancers. Results: Fourteen immune-related metabolic signatures (FABP6, LPA, RBP4, CETP, STAB2, PPARG, TYMP, CGA, GCGR, SDC1, BGN, ABCA1, PLA2G4A, PLK1) were identified for use in constructing a comprehensive prognostic model for BC. The high-risk group was characterized by poorer diagnosis, fewer activated immune cell infiltration and better treatment response to immune checkpoint inhibitors. Moreover, the index was combined with clinical parameters, weighted, and created a nomogram. It is imperative to point out that our model and corresponding nomogram are optimal and independent prognosis factors compared to other traditional clinical variables. They also have satisfactory predictive capacity validated by ROC curve, calibration plot and DCA analysis. Conclusions: Our 14-MRDEGs and their multiple integrations reflected genetic-level and immunometabolic profile alterations in BC, allowing accurate prediction of survival risk and the efficacy of immunotherapy. The research conclusions may provide a reference for further analysis and drug development in target discovery.