Abstract 7479: Single-cell OncoTreat: Novel therapeutic candidates by high resolution inference and validation of drug sensitivities in heterogeneous cholangiocarcinoma tumors

Abstract Cholangiocarcinoma is an aggressive biliary adenocarcinoma with limited treatment options. Stromal infiltration and intratumoral heterogeneity in this tumor type prevents identification of actionable dependencies from bulk-tissue profiles. While cholangiocarcinoma illustrates these challenges, they are common across many other aggressive tumors.To address these challenges, we introduce a single-cell RNA-Sequencing (scRNA-Seq) framework for prioritization of drugs to treat heterogeneous tumor cell subpopulations. Single-cell analyses can bypass limitations of sample size in a precision medicine approach by providing large tumor cell “cohorts” even from individual patients, with the advantage that cell subpopulations can be effectively distinguished. Unfortunately, >80% of genes in each cell profiled by scRNA-Seq are typically undetectable, limiting discoverability of therapeutic targets and drug sensitivities. The VIPER algorithm addresses this by inferring the transcriptional activity of a protein from expression of its downstream target genes. By then identifying aberrantly activated proteins, each cell can be matched to an empiric database of cell line drug perturbation RNA-Seq, ranking drugs which empirically inhibit the proteins aberrantly activated in each cell. We refer to this approach as single-cell OncoTreat.Analysis of tumor cells across six patients by this approach has revealed three distinct sub-phenotypes of cholangiocarcinoma cells, with conserved proteomic signatures, and distinct predicted drug sensitivities. A seventh patient recapitulated these phenotypes, with drug predictions validated in a low-passage, patient-derived xenograft (PDX) model. The top-ranked drug by integrated analysis of all tumor cells was plicamycin, which at sub-cluster level only two of three clusters were predicted to respond to. Conversely, dacinostat was inferred to be active against all three tumor-cell sub-phenotypes. Tumor growth studies in PDX demonstrated growth rate control with both plicamycin and dacinostat, but much improved survival time with dacinostat, outperforming gemcitabine + cisplatin standard of care. Post-treatment scRNA-Seq confirmed predicted subpopulation-specific effects, such that dacinostat depleted all three tumor sub-phenotypes, while plicamycin depleted only the two predicted, allowing the third to reconstitute a treatment-resistant tumor. To our knowledge, this is the first report of dacinostat as an HDAC inhibitor effective in cholangiocarcinoma treatment, in vivo. Ultimately, this approach suggests promising candidates for follow-up trials in cholangiocarcinoma, and is highly generalizable to elucidate complementary dependencies of other aggressive, heterogeneous tumors. Citation Format: Aleksandar Obradovic, Lorenzo Tomassoni, Daoqi You, Charles Drake, Filemon Dela Cruz, Andrew Kung, Andrea Califano. Single-cell OncoTreat: Novel therapeutic candidates by high resolution inference and validation of drug sensitivities in heterogeneous cholangiocarcinoma tumors [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 7479.