Constructing a prognostic signature of tumor-associated B lymphocytes in hepatocellular carcinoma via machine learning integration.

B lymphocytes are known for their anti-tumor role in hepatocellular carcinoma (HCC), but recent studies suggest that tumor-tamed B cells can promote cancer progression in other malignancies. It remains unclear whether a similar subset exists in HCC. We analyzed single-cell RNA RNA-seq (GSE162616) to compare B lymphocytes in normal and HCC samples. Weighted correlation network analysis (WGCNA) identified tumor-associated B lymphocyte module genes. Furthermore, we utilized 101 machine learning algorithms to construct a prognostic model, which was further compared with other published prognostic models using statistical comparisons, including P-values and confidence intervals. The model was further tested for its ability to predict responses to HCC-specific immunotherapy and chemotherapy in available public datasets; non-HCC cohorts were exploratory. A distinct subset of B lymphocytes was significantly elevated in HCC, exhibiting increased IgG secretion and TGFBR1 expression. Functional validation experiments confirmed that siRNA-mediated knockdown of SOX4 and TGFBR1 significantly suppressed HCC cell proliferation, supporting their tumor-promoting roles. WGCNA identified key module genes associated with these tumor-associated B cells. The RSF + plsRcox model generated the tumor-associated B lymphocyte score (TABLS), which demonstrated superior predictive performance for overall survival (OS) and relapse-free survival (RFS) compared to other prognostic models. Patients with low TABLS were more likely to respond favorably to immunotherapy and chemotherapy. External validation was performed in publicly accessible HCC cohorts with individual-level outcomes; prospective validation in additional HCC-specific immunotherapy cohorts will further strengthen clinical relevance. TABLS is a promising prognostic biomarker for HCC, with potential clinical applications in therapy selection.