RTK-RAS pathway alterations in hepatocellular carcinoma and cholangiocarcinoma: a multicohort genomic and single-cell...

To determine whether heterogeneous genomic alterations detected by clinical tumor panels converge on a reproducible pathway-level difference between hepatocellular carcinoma (HCC) and cholangiocarcinoma (CCA), and to examine its transcriptional output, cellular localization, and tumor-stromal coexpression context in public multi-omics cohorts.

The institutional cohort comprised 52 patients who underwent tumor tissue genomic testing, including 28 with HCC and 24 in a CCA working cohort whose intrahepatic location was not uniformly documented. Ten canonical oncogenic pathways were compared using panel-aware evaluable denominators, and the selected pathway was tested in independent Memorial Sloan Kettering (MSK) cohorts. Matched genomic and transcriptomic data from The Cancer Genome Atlas (TCGA) were used to evaluate downstream pathway activity. GSE151530 and GSE189903 were analyzed by patient-level pseudobulk to localize the transcriptional difference, followed by CellChatDB-guided ligand-receptor coexpression analysis.

Receptor tyrosine kinase-RAS (RTK-RAS) showed the strongest cancer-type difference in the institutional cohort: alterations were detected in 4/21 evaluable HCC cases and 8/11 evaluable CCA cases (odds ratio [OR] = 9.44; 95% CI, 1.87-47.70; P = 0.00601; Benjamini-Hochberg-adjusted q = 0.0601), with concordant directions across sensitivity analyses. In MSK, integrated RTK-RAS alterations occurred in 240/412 intrahepatic cholangiocarcinoma (iCCA) cases (58.3%) and 107/289 HCC cases (37.0%; OR = 2.37; 95% CI, 1.74-3.22; P = 3.12 × 10^-8). In TCGA, RTK-RAS status was associated with higher mitogen-activated protein kinase (MAPK) activity in the pooled cancer-type-adjusted model (β = 0.240; 95% CI, 0.051-0.428; P = 0.0129), whereas the within-iCCA estimate was positive but imprecise. Patient-level single-cell meta-analysis localized the higher iCCA MAPK activity predominantly to malignant cells (pooled effect = 1.455; 95% CI, 0.658-2.252; P = 3.46 × 10^-4; q = 0.00208). Platelet-derived growth factor (PDGF)-PDGFRA and insulin-like growth factor (IGF)-IGF1R coexpression patterns accompanied this malignant-cell phenotype.

Panel-aware integration of clinical genomic testing and public multi-omics data supports a higher burden of heterogeneous RTK-RAS alterations in CCA/iCCA than in HCC and links this difference to a malignant-cell MAPK transcriptional phenotype. The pathway-level framework complements, rather than replaces, established alteration-specific interpretation and generates tumor-stromal coexpression hypotheses for further spatial and functional testing.