Abstract
OBJECTIVES: In this study, we constructed a model based on circadian rhythm associated genes (CRRGs) to predict prognosis and immune infiltration in patients with breast cancer (BC). MATERIALS AND METHODS: By using TCGA and CGDB databases, we conducted a comprehensive analysis of circadian rhythm gene expression and clinicopathological data. Three different machine learning algorithms were used to screen out the characteristic circadian genes associated with BC prognosis. On this basis, a circadian gene prediction model about BC prognosis was constructed and validated. We also evaluated the association of the model's risk score with immune cells and immune checkpoint genes, and analyzed prognostic genes and drug sensitivity in this model. RESULTS: We screened 62 DEGs, including 30 upregulated genes and 32 downregulated genes, and performed GO and KEGG analysis on them. The above 62 DEGs were included in Cox analysis, LASSO regression, Random Forest and SVMV-RFE, respectively, and then the intersection was used to obtain 5 prognostic related characteristic genes (SUV39H2, OPN4, RORB, FBXL6 and SIAH2). The Risk Score of each sample was calculated according to the expression level and risk coefficient of 5 genes, Risk Score= (SUV39H2 expression level ×0.0436) + (OPN4 expression level ×1.4270) + (RORB expression level ×0.1917) + (FBXL6 expression level ×0.3190) + (SIAH2 expression level × -0.1984). CONCLUSION: SUV39H2, OPN4, RORB and FBXL6 were positively correlated with Risk Score, while SIAH2 was negatively correlated with Risk Score. The above five circadian rhythm genes can construct a risk model for predicting the prognosis and immune invasion of BC.