Three-gene neural model distinguishes tumor from adjacent liver tissue arrays
Source check: Source checked. Not independent scientific confirmation.
Finding: A learned three-gene model distinguished known HBV-associated liver-tumor tissue from adjacent tissue in public expression datasets. The source reports validation AUC .849 across 140 arrays; the original dataset includes multiple samples from 11 patients. This is tissue classification, with no evidence of early screening or future cancer conversion.
Technology: Random forest, SVM-RFE, LASSO and three-gene artificial neural network
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- Learned feature-selection algorithms and an ANN discriminate tissue labels; conventional enrichment and inferred immune scores are not separately counted as new experimental biology.
- Test and result
- Intersecting RF, SVM-RFE and LASSO selections produced HHIP, CXCL14 and CDHR2. The source used GSE121248 with 70 tumor/37 adjacent arrays for development and GSE55092 with 49 tumor/91 adjacent arrays for reported validation AUC .849(95% CI .773–.916), compared with training AUC .948. Original GEO metadata describes up to 17 whole-liver tissue samples per liver from 11 HCC patients, with selected laser-capture samples also included.
- Limitation / still unknown
- Array counts are not independent patient counts. Adjacent tissues come from livers already affected by HCC, not a prospective cohort of people with uncomplicated hepatitis B. Multiple regions and sample modalities can affect discrimination and uncertainty. GeneScore median thresholds were computed on each dataset, so the validation transformation is not a demonstrated frozen deployment rule. No blood test, prospective early detection, future HBV-to-cancer conversion, causal immune mechanism or clinical benefit was validated. The source association/subtype analyses reuse these datasets.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10638043/
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: linked separately.
Review date: 21 Nov 2026. Review scope: Original data definition, learned feature screening, ANN Methods and reported validation Results actually read; original GEO indexed study-design metadata independently checked..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 1/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.