Shared Expression Signals in Liver Cancer

Evidence

This integrated bioinformatics study examined publicly available human hepatocellular carcinoma (HCC) datasets spanning HBV-associated, HCV-associated, and non-viral HCC. The workflow combined gene-expression analysis, protein-protein interaction network construction, functional enrichment, hub-gene identification, database validation, and survival analysis. The analyzed microarray datasets were GSE19665, GSE121248, GSE44074, GSE17856, GSE62232, and GSE45050. Candidate genes were evaluated in GEPIA using TCGA and GTEx data, with logFC and p-values used as cut-off criteria. The study therefore compared molecular expression patterns among etiologic HCC groups and then assessed selected candidates against normal liver tissue and survival data, rather than testing a therapeutic or diagnostic intervention. pmid:42993760

For database validation, the authors examined candidate genes in LIHC relative to normal liver tissue in TCGA data. They reported significant upregulation of GPC3, SPARCL1, LEF1, THBS4, and VWF, alongside downregulation of ADAMTS13. TNMplot analyses of gene and protein expression were reported to align with the study data. The TNMplot validation compared Normal tissue (n = 376) with Tumor tissue (n = 806). This is a tissue-expression comparison: it describes differences between the analyzed normal and tumor liver samples and does not measure a change caused by altering any of these genes in patients. The result is particularly relevant to a proteins and proteomics framing because TNMplot was used for both gene- and protein-expression analysis, while the broader analytic workflow also used protein-protein interaction networks to prioritize hub genes. pmid:42993760

The reported logFC table showed that GPC3 was positive across the HBV, HCV, HCC, and GEPIA columns, and SPARCL1 was likewise positive across those columns. These tabled patterns were part of the authors’ effort to identify common molecular features spanning viral and non-viral liver cancer. Functional enrichment and network analyses linked the candidate genes to pathways and processes involved in HCC progression. The study also reported SPARCL1 as a potential inhibitor of SPP1. That characterization arose from the integrated computational analysis and should remain distinct from a demonstrated biochemical inhibition experiment in a defined laboratory system. pmid:42993760

The survival component used the TCGA-LIHC cohort through GEPIA. Patients were stratified into high- and low-expression groups using the median expression value, described as a 50% cutoff. Kaplan–Meier analysis assessed the association between high versus low expression of GPC3, SPARCL1, THBS4, LEF1, VWF, and ADAMTS13 and overall survival. The report describes a population of 182 people in both expression groups. The source reports correlations between expression level and overall survival for these genes, but the stated design does not establish that expression of any individual candidate causes a survival difference. The comparison is an expression-stratified association in a database cohort, and it is not an evaluation of protein-targeted treatment. pmid:42993760

Analysis — Integrated expression pattern

The central pattern is convergence rather than a single-gene test: candidate-gene signals were assembled across HBV-associated, HCV-associated, and non-viral HCC datasets, then examined through network, pathway, tissue-expression, protein-expression, and survival analyses. GPC3, SPARCL1, LEF1, THBS4, VWF, and ADAMTS13 form the most directly validated set because the study explicitly compared their expression in LIHC versus normal liver tissue and linked high-versus-low expression groups to overall survival. Within this design, agreement between the study’s data and TNMplot is a cross-database consistency observation, not independent experimental confirmation of protein function. Similarly, the association of gene-expression strata with overall survival provides a prognostic pattern in the TCGA-LIHC analysis, not evidence that modifying expression would change outcomes. The selection process—pooling datasets within etiology groups, identifying shared hub genes, and validating selected candidates in external databases—frames the findings as a computationally prioritized shared-expression signature. The data are consequently most informative for molecular characterization of the analyzed human HCC tissue datasets and for defining candidates whose functions require direct testing. pmid:42993760

Limitations

The analysis relied on publicly available datasets and lacked experimental validation. Within each etiology group, the study pooled multiple GEO datasets generated on different microarray platforms before comparing the resulting pooled groups. The authors state that this approach may introduce inter-dataset heterogeneity and potential batch effects; no explicit batch-correction method was applied before pooling. These features constrain the robustness of comparisons across HBV-associated, HCV-associated, and non-viral HCC groups because measured differences can reflect both biological variation and differences among contributing datasets. Clinical and demographic metadata—including age, sex, disease stage, tumor grade, and treatment history—were not uniformly reported by the original GEO submitters, precluding refined subgroup analyses based on those variables. Finally, tissue-expression validation, protein-expression database concordance, and survival associations do not replace in vitro or in vivo validation of candidate-gene function. pmid:42993760