Integrating Immunogenetics, Single-Cell Transcriptomics, and AI-Assisted Drug Discovery Reveals CXCL10-Associated Immune Remodelling in Breast Cancer

Breast cancer is a heterogeneous malignancy shaped by genetic alterations and cytokine-mediated immune interactions. However, integrative characterization of cytokine-related regulators using genetic, single-cell, and computational approaches remains limited. This study aimed to identify genetically supported cytokine-related candidates in breast cancer, characterize their immune and cellular contexts, and explore their potential regulatory and therapeutic relevance. Bulk transcriptomic datasets were integrated to identify differentially expressed cytokine-cytokine receptor interaction-related genes, followed by two-sample Mendelian randomization (MR), functional enrichment, immune infiltration, single-cell RNA sequencing, CellChat analysis, in silico gene perturbation, AI-assisted compound screening, and molecular docking. A total of 882 differentially expressed genes and 20 cytokine-related candidates were identified. MR provided the most consistent support for CXCL10, with genetically predicted higher expression associated with increased breast cancer risk (OR = 1.112, 95% CI: 1.026-1.206, p = 0.010). PRLR was significant only in the IVW analysis and was therefore considered a candidate requiring further validation. The CXCL10/PRLR model achieved an AUC of 0.810 (95% CI: 0.774-0.846) for discriminating breast cancer from normal samples. CXCL10 was predominantly associated with immune-cell populations, interferon-related programs, and predicted chemokine communication. In silico perturbation, AI-assisted screening, and docking further generated computational hypotheses regarding CXCL10-associated regulation and candidate compounds. CXCL10 represents a genetically supported immune-associated candidate in breast cancer. Overall, the findings are computational and hypothesis-generating and require validation in independent cohorts and experimental models.