Work

Weighted gene co-expression network of tumor and normal liver transcriptomes identifies prognostic signature in hepatocellular carcinoma

Year: 2026
Type: article
Abstract: BACKGROUNDMany transcriptome-based signatures for hepatocellular carcinoma (HCC) prognosis have been reported, but most show limited portability across cohorts.We hypothesized that a network-guided analysis of harmonized liver transcriptomes could identify a biologically coherent and portable prognostic signature. AIMTo identify biologically coherent co-expression modules and develop a robust expression signature for HCC prognosis. METHODSWe used uniformly processed RNA-seq data from the University of California Santa Cruz Toil recompute compendium to integrate The Cancer Genome Atlas (TCGA)-liver HCC tumors (n = 371), TCGA adjacent non-tumor liver (n = 50), and Genotype-Tissue Expression normal liver (n = 110).After filtering out low-expression genes and selecting 8000 most variable ones according to their median absolute deviation, we performed weighted gene co-expression network analysis to identify co-expression modules in tumor and normal tissues, which was followed by hub-gene prioritization, enrichment analysis, least absolute shrinkage and selection operator-Cox modeling, external validation, and meta-analysis. RESULTSWeighted gene co-expression network analysis identified a tumor-associated module enriched for cell-cycle progression, mitotic checkpoint control, DNA replication, E2F targets, and mTORC1 signaling.A normal-liver-associated module showed a distinct expression module enriched for detoxification, metal-ion/toxin response, and liver 5 / 51 homeostatic functions.Cross-platform filtering, survival-analysis screening and least absolute shrinkage and selection operator-Cox modeling yielded a 3-gene signature consisting of MMP1, KIF2C, and HOXD10.In TCGA, the signature-based risk score was significantly associated with overall survival, and the association remained significant after adjustment for age at diagnosis, sex, and disease stage (hazard ratio = 2.95, 95% confidence interval: 1.92-4.53,P = 8.10 × 10 -7 ).The TCGA-derived coefficients were then applied to six external cohorts, with adverse prognostic associations observed in the same direction across all cohorts, although statistical significance varied among individual datasets.Random-effects meta-analysis yielded a pooled hazard ratio of 1.54 (95% confidence interval: 1.30-1.82)per 1-SD increase in risk score (I 2 = 39.8%). CONCLUSIONThe network-guided analysis of a harmonized TCGA-Genotype-Tissue Expression liver transcriptome identified a biologically coherent survival-related expression signature with cross-platform prognostic associations in retrospective cohorts.This signature represents a candidate prognostic model requiring prospective assay-based validation and evaluation of clinical utility.
Language: English
Subfield: Hepatology
Field: Medicine
Open Access status: diamond
OpenAlex ID: W7225994297
FWCI:
Cites: 42
Cited by:
Related to: 10