Abstract: Objective To predict the metastasis risk of lymph node-negative breast cancer by combination gene expression profiles with clinical and pathological data. Methods The samples with the characteristics of radical mastectomy,primary tumor ≤5 cm,lymph node-negative and non-neoadjuvant treatment were searched from GEO database.They were randomly divided into training and validation group at ratio of 2∶1.The distant metastasis free survival (DMFS) closely related genes in the training group were selected to fit Cox regression model with age,histological type,tumor size,and endocrine treatment.The risk of distant metastases for the validation group was predicted from the Cox regression model,and the survival rate was calculated by Kaplan-Meier survival analysis and Log rank test. Results Six thousand two huandred and fourteen gene expression profiles of 367 cases were included.Thirty two gene expression levels were closely related with distant metastasis free survival(DMFS)(P<0.001) in training group,prognostic index was calculated according to the COX proportional hazards model in validation group of 122 patients,the median DMFS was (4 026±382) days in the high-risk group(prognostic index>0.833) and (7 237±382) days in the low -risk group(prognostic index ≤0.833),the difference was statistically significant (χ²=5.900，P=0.015). Conclusion It is feasible and effective to predict the metastasis risk of lymph node-negative breast cancer by combination 32 gene expression profiles with clinical and pathological data.