Objective To explore the disease targets of breast cancer related to age at diagnosis and tumor stage by weighted gene co-expression network analysis (WGCNA) from public database The Cancer Genome Atlas (TCGA). Methods We obtained the breast cancer gene chip expression data and corresponding clinical data of 53 Asians and 126 Africans from TCGA database. R software WGCNA package was used to construct the co-expression network of the two populations, and the significant modules related to age at diagnosis and cancer stage were obtained. Online website DAVID was used for function enrichment and online website UALCAN for survival analysis. Results WGCNA yielded 11 modules significantly related to cancer stage and age at diagnosis. Forty-two candidate genes were obtained after 11 modules were intersected. Gene ontology (GO) enrichment analysis was carried out using online website DAVID and these genes were mainly involved in protein binding function. Nine of the 42 candidate genes were identified as hub genes by WGCNA, the 9 genes were used in UALCAN for differential analysis and survival analysis, and 2 candidate biomarkers (ERLIN2 and ASH2L) were screened out. The expression of the 2 genes in normal tissues and breast cancer tissues was significantly different (P<0.01), and the expression level significantly influenced the survival of breast cancer patients (P<0.05). Conclusion Data mining from public databases for biomarkers or therapeutic targets is a cost-effective research method. In this study ERLIN2 and ASH2L have been found to be candidate biomarkers for breast cancer through data mining, which needs large sample study and mechanism exploration.