Analysis of Stemness and Prognosis of Subtypes in Breast Cancer Using the Transcriptome Sequencing Data

Abstract

The stem characteristics of tumor cells have been proposed in theory very early, and we can use the signature of gene expression to speculate the stemness of tumor cells. However, systematic studies on the stemness of breast cancer as well as breast cancer subtypes, and the relationship between stemness and metastasis and prognosis, are still lacking. In the present research, using the transcriptome data of patients with breast cancer in the TCGA database, a stemness prediction model was utilized to derive the stemness of the patients’ tumors. We compared the stemness values among different subtypes and the differences with metastasis. COX regression was employed to evaluate the correlation between stemness value as well as prognosis. Using the Lasso-penalized Cox regression machine learning model, we obtained the gene signature of the basal subtype that is related to stemness and can also predict the prognosis of the patient. Patients can be stratified into two groups of high and low stemness, corresponding to good and poor prognosis. Based on further prediction of tumor infiltration by CIBERSORT and prediction of drug response by a connectivity map, we found that the difference in stemness between these two groups is associated with the activation of tumor-killing immune cells and drug response. Our findings can promote the understanding and research of subtypes of basal breast cancer and provide corresponding molecular markers for clinical detection and therapy.