Identification Of Breast Cancer Subtypes By Intergrated Genome And Immune Microenvironment

Objective: The aim of this study was to identify breast cancer(BC)molecular subtypes and construct prognostic models based on key gene classifiers by combining genomic analysis with tumor immune microenvironment(TME).Method: Download the open data set of Cancer Genome Atlas(TCGA)for unsupervised consensus clustering to obtain the type clusters of BC patients,and identify the type according to the expression profile.The prognosis and enrichment pathways of each type were compared.Then,samples were extracted from the open dataset of Gene Expression Omnibus(GEO)and the Molecular Classification of Breast Cancer International Consortium(METABRIC)dataset to verify the correlation between typing and prognosis.Weighted gene co-expression network analysis(Weighted gene Coexpression Network Analysis,WGCNA)was used to select the gene module with the greatest correlation with typing,and the minimum absolute shrinkage selection operator(Lasso)was used to extract the key typing classification label,a prognostic model of categorical labels was established by Cox regression analysis,the results showed that according to the median of the risk score to divide high and low risk,the high risk group was associated with poorer overall survival(OS)compared with the low risk group closely related.Subsequently,we performed validation in the METABRIC dataset,univariate and multivariate regression analysis to obtain the typing classification labels and their biological functions,and this set of labels was validated by the BC dataset of the Metabric database.In addition,single-cell analysis was performed to assess the association between BC classification and TME.Finally,drug sensitivity and immune cell infiltration in different phenotypes of BC patients were also calculated by CIBERSORT and ESTIMATE algorithms.Result: TCGA BC samples were divided into three subgroups with significantly different prognosis,S1,S2,and S3,in which S2 had a poorer prognosis,and S1 and S3 had a better prognosis.Three key pathways and ten key prognosis-related gene signatures were screened by enrichment analysis and cox regression analysis.Finally,single-cell analysis revealed that S1 samples had the most types of immune cells,S2,which was more sensitive to tumor therapeutic drugs,had more neutrophil enrichment,and more polylymphoid progenitor cells were involved in typing S3.Conclusion: Our novelty lies in the identification of BC typing and typing taxonomic labels using a large dataset combined with multiple bioinformatics approaches.The research results provide a basis for clinical precision treatment of BC.