Triple-negative breast cancer (TNBC) is defined as a kind of breast cancer that lacks estrogen receptors (ER), progesterone receptors (PR), and human epidermal growth factor receptors (HER2). This cancer accounts for 10–15% of all breast cancers and has the features of high invasiveness and metastatic potential. The treatment regimens are still lacking and need to develop novel inhibitors for therapeutic strategies. Three-dimensional quantitative structure-activity relationship (3D-QSAR) analyses, based on a series of forty-seven thieno-pyrimidine derivatives, were performed to identify the key structural features for the inhibitory biological activities. The established comparative molecular field analysis (CoMFA) presented a leave-one-out cross-validated correlation coefficient q2 of 0.818 and a determination coefficient r2 of 0.917. In comparative molecular similarity indices analysis (CoMSIA), a q2 of 0.801 and an r2 of 0.897 were exhibited. The predictive capability of these models was confirmed by using external validation and was further validated by the progressive scrambling stability test. From these results of validation, the models were determined to be statistically reliable and robust. This study could provide valuable information for further optimization and design of novel inhibitors against metastatic breast cancer.