Hepatocellular carcinoma (HCC) is characterized by coordinated transcriptional and post-transcriptional dysregulation. We sought to identify clinically relevant miRNA–mRNA regulatory axes with cross-cohort, multi-layer validation. Tumor–normal differential expression was integrated across TCGA-LIHC and independent microarray cohorts. Predicted miRNA targets were filtered by inverse-direction overlap with consensus DEGs and evaluated for activity–abundance coherence and survival relevance. An eight-feature axis (four miRNAs and four target-set activities) was modeled using penalized Cox regression with cross-validation and externally validated. Four recurrent miRNAs were identified (miR-125b-2 downregulated; miR-21, miR-221, miR-9-1 upregulated). miR-125b-2 showed the strongest inverse coherence with its target set (ρ = −0.41, P < 1 × 10⁻¹²). The composite axis stratified TCGA overall survival (log-rank P < 0.0001) and validated in GSE31384 (HR = 1.62, 95% CI 1.15–2.28; C-index = 0.66). Seven targets met FDR ≤ 1%, with UCK2 exhibiting the strongest adverse association (HR = 2.78, P = 2.7 × 10⁻⁹). UCK2 was overexpressed, hypomethylated, enriched in epithelial compartments, and linked to proliferative and ECM/growth-factor signaling programs. Functional assays demonstrated that UCK2 knockdown suppressed proliferation, clonogenicity, and migration, whereas overexpression enhanced these phenotypes. The miR-125b-2/UCK2 axis defines a metabolically driven, epithelial proliferative program coupled to ECM/GF signaling and represents a validated prognostic and biologically actionable node in HCC.
Zhao et al. (Thu,) studied this question.