Single‐Cell Multi‐Omics Reveals <scp>B2M</scp> ‐Mediated Myeloid Reprogramming and Constructs a Predictive Model for Early Hepatocellular Carcinoma Recurrence

Early recurrence remains a major challenge in the management of hepatocellular carcinoma (HCC), yet its molecular mechanisms are not fully understood. In this study, we applied an integrative multi-omics strategy at single-cell resolution to explore potential drivers of early HCC recurrence (recurrence time < 2 years) and to develop a predictive framework. By combining single-cell RNA sequencing, proteomics, transcriptomics, and clinical feature analysis, we identified 14 relapse-associated proteins, including CD274, B2M, MYC, and CASP3, as candidate risk factors. Transcriptomic profiling suggested the enrichment of pathways such as MYC-TARGETS-V2 and INTERFERON-GAMMA-RESPONSE. Single-cell analysis indicated reduced immune cell infiltration in recurrent tumors, with myeloid cells (particularly cDC2 and macrophages) showing B2M-associated reprogramming characterized by HLA downregulation and altered GAS6/PROS1 signaling, consistent with tumor-associated macrophage-like phenotypes. A LASSO regression model based on cDC2 and macrophage signature genes demonstrated moderate predictive performance in both the training and validation cohorts (AUC > 0.65). Drug sensitivity analyses further suggested that vandetanib may have the potential to inhibit recurrence by targeting B2M-related pathways. These findings provide evidence that B2M may contribute to remodeling of the immune microenvironment in recurrent HCC. Our integrative single-cell multi-omics approach highlights a possible mechanism of early recurrence and offers a preliminary predictive tool with therapeutic implications.