Objective To identify candidate bifunctional drugs that simultaneously modulate molecular features associated with primary triple-negative breast cancer (TNBC) and bone metastasis based on an artificial intelligence (AI)-driven strategy integrating driver gene identification and transcriptomic differentially expressed gene analysis. Methods Based on the Fudan University Shanghai Cancer Center (FUSCC) TNBC cohort and Cancer Cell Line Encyclopedia (CCLE) data, after eliminating the aberrant samples through expression correlation analysis, 2 comparison groups were established according to the research objectives: bone metastasis tissue group (BMTG), including tumor tissue from patients with bone metastasis (n＝17) and the adjacent normal tissue (n＝6); primary tumor group (PTG), including patients with bone metastasis (n＝17) and those without bone metastasis (n＝271) among primary tumor samples. In the FUSCC TNBC cohort, potential driver genes were identified using PhenoDriver in the BMTG and differentially expressed genes were obtained using differential gene analysis in the PTG, and the intersection of both was taken with Landmark genes from the Library of Integrated Network-Based Cellular Signatures (LINCS) 2020 to establish disease-specific gene sets. BMTG was iteratively screened and evaluated by receiver operating characteristic (ROC) curve analysis using the FUSCC TNBC cohort, The Cancer Genome Atlas (TCGA)-TNBC and 3 Gene Expression Omnibus (GEO) datasets; while PTG was validated by ROC curve analysis using FUSCC TNBC cohort. Robust core disease gene sets were obtained for drug reversal prediction. Subsequently, connectivity scores were calculated based on the Connectivity Map (CMap), and candidate bifunctional drugs were screened using a comprehensive ranking metric. Combined with previous literatures and commercial availability of compounds, SB-590885 and PF-431396 were prioritized for cell experimental validation. TNBC cell line MDA-MB-231 and TNBC bone metastatic cell line MDA-BoM-1833 were used to verify the inhibitory activities of the 2 candidate drugs by cell counting kit 8 assay and cell wound-healing assay. Results A total of 72 aberrant samples were excluded. Ultimately, 26 BMTG driver genes and 62 PTG differentially expressed genes were identified as the core disease gene sets. Based on drug repositioning of LINCS perturbation profiles, 46 potential bifunctional drugs were screened. SB-590885 and PF-431396 demonstrated significant dose-dependent inhibitory effects on the proliferation and migration of TNBC cell lines and TNBC bone metastatic cell lines in vitro. Conclusion This study has identified SB-590885 and PF-431396 as potential bifunctional therapeutic candidates, providing new research directions and potential treatment options for TNBC with bone metastasis.