Transcriptomic drug repositioning in hepatocellular carcinoma guided by disease signatures and tiered translational evidence

Systemic therapy for hepatocellular carcinoma (HCC) has advanced rapidly. However, durable benefit remains constrained by resistance, tumor heterogeneity, impaired liver function and weak predictive biomarkers. This narrative review examines how HCC disease signatures are constructed, queried against CMap/LINCS/L1000 perturbational resources and translated into prioritized compounds. We map representative studies by signature source, perturbational platform, candidate output, validation depth and principal limitations. To reduce overinterpretation, we propose a claim-calibrated framework, defined here as an author-proposed, non-validated five-tier interpretive model that aligns the strength of candidate language with the depth of supporting evidence. The tiers distinguish single computational nomination, recurrence across signatures or methods, target/pathway plausibility, direct HCC validation and translational readiness. Connectivity scores are therefore treated as prioritization signals rather than evidence of efficacy. A SIB-1893/GRM5-linked case was selected because it combines one documented HCC-related computational nomination with target-level plausibility while lacking compound-specific HCC validation. It therefore illustrates why target-level context does not establish independent recurrence or direct HCC activity. We also compare transcriptomic reversal with network-based, AI-driven, multimodal and foundation-model approaches and discuss how etiology, single-cell resolution and spatial context may refine future signatures. Overall, transcriptome-guided repositioning is best used as a hypothesis-generating strategy; therapeutic relevance should be claimed only after robust signature construction, transparent reporting, liver-relevant pharmacological filtering and direct HCC-specific validation.