Molecular classification of biliary tract cancer by RNA clustering analysis and its clinicopathological characteristics.

Abstract

143 Background: Biliary tract cancer (BTC) is a rare malignancy but occurs more frequently in Asian populations, where the establishment of molecularly guided treatment strategies is of particular importance. Intrahepatic cholangiocarcinoma has been classified into small-duct and large-duct types based on molecular features, each characterized by distinct genetic alterations. However, the clinical significance of RNA expression–based molecular classification in BTC remains unclear. Methods: Patients with BTC enrolled in the multinational MASTER KEY Asia registry, including the CHOICE sub-cohort, who had matched DNA sequencing and RNA expression data by NGS were included. Gene expression profiles from 93 cholangiocarcinoma tumors were analyzed using unsupervised hierarchical clustering to define transcriptomic subgroups, and heatmaps were generated using scaled expression values. Mutation profiles were aligned with clusters, and the top 10 most frequently mutated genes in each group were compared. Clinicopathological features and treatment responses were evaluated across subgroups. Overall survival (OS) and progression-free survival (PFS) were assessed using Kaplan–Meier and log-rank tests, with Cox proportional hazards and logistic regression analyses performed to identify prognostic factors and determinants of cluster assignment. Results: Unsupervised clustering identified two major transcriptomic groups, designated CHOL1 (n=61) and CHOL2 (n=32), with distinct gene expression profiles. The median age was significantly lower in CHOL2 compared with CHOL1 (56.3 vs. 62.7 years, p=0.009). In addition, logistic regression analysis demonstrated that younger age and a positive family history of cancer were significantly associated with CHOL2 membership (p=0.013 and p=0.009, respectively). Distinct genetic alteration patterns were identified: KRAS (36.1% vs. 9.4%, p=0.006) and SMAD4 mutations (26.2% vs. 3.1%, p=0.005) were more frequent in CHOL1, whereas FGFR2 fusions (1.6% vs. 9.4%, p=0.11) and IDH1 mutations (0% vs. 15.6%, p=0.003) were numerically more common in CHOL2. In CHOL2, a subset of cases harbored TERT promoter mutations and exhibited gene expression signatures associated with aggressive HCC. However, no significant differences were observed in response rates to systemic therapy, PFS, or OS between clusters. Conclusions: RNA-based molecular classification of BTC identified two distinct clusters with different clinical and genomic characteristics. CHOL2 was associated with younger age and enrichment of actionable alterations such as FGFR2 fusions and IDH1 mutations. Although no differences in treatment efficacy were observed between clusters, heterogeneity in treatment regimens across countries may have limited evaluation, highlighting the need for further validation in homogeneous cohorts. Clinical trial information: NCT05217407 .