Tumor-Intrinsic Hepatocyte Arm-Level Genomic States Shape Immunotherapy Response Heterogeneity in Hepatocellular Carcinoma

Purpose: Bulk transcriptomic biomarkers for immune checkpoint inhibitor (ICI) response in hepatocellular carcinoma (HCC) often lack reproducibility because bulk RNA sequencing captures composite signals from malignant, immune, and stromal compartments. Variability in tumor purity and malignant cell composition can confound immune-based interpretations. We developed an integrative framework combining single-cell-derived digital cytometry with inference of tumor-intrinsic genomic states to better interpret transcriptomic variation associated with ICI response. Materials and Methods: Single-cell RNA sequencing data from HCC tumors (GSE206325) were used to construct a nine-cell-type signature matrix for CIBERSORTx deconvolution and to infer chromosome arm-level copy number variation in malignant hepatocytes using inferCNV. Signature stability was evaluated through pseudobulk reconstruction and gradient simulations. Digital cytometry was applied to three bulk RNA-seq ICI cohorts (GSE202069, GSE215011, and GSE279750). Arm-level alterations were projected onto bulk transcriptomes by mapping arm-associated genes, standardizing expression within samples, and aggregating direction-adjusted Q90 statistics into a composite arm-axis score. Results: Digital cytometry revealed cohort-dependent variability in malignant hepatocyte dominance and limited reproducibility of immune fraction differences. Differential expression analysis also showed poor cross-cohort concordance. InferCNV identified recurrent arm-level alterations (1q/8q gain, 12p/13q loss), defining a continuous genomic axis. In pooled analysis (n=36), integrating the arm-axis score with PD-L1 improved discrimination (AUC 0.775; 95% CI, 0.602-0.920). In TCGA-LIHC (n=361), arm-level burden was inversely associated with cytolytic activity and positively associated with proliferation. Conclusion: Tumor-intrinsic hepatocyte genomic states provide complementary predictive information beyond immune activation alone and may help explain heterogeneity in ICI response across HCC cohorts.