Single-cell RNA-seq (scRNA-seq) is an emerging platform for high-throughput profiling of individual cells in a sample and is routinely employed to investigate the transcriptional landscapes of the cellular constituents of tumors. To this end, many scRNA-seq specific clustering algorithms have emerged to analytically partition cells into modules of comparatively similar profiles. To facilitate data driven molecular subtyping of such scRNA-seq clustering results and other large-scale-omics studies, we have developed K2 Taxonomer. K2 Taxonomer is an R package built around a novel top-down hierarchical clustering algorithm, utilizing repeated perturbations of the data to generate robust taxonomical partitions of observations. The software runs additional analyses to define gene co-expression signatures of these modules, as well as to integrate user-input annotations of genes and/or observations. An interactive web portal has been generated to assist in the interrogation of the full compendium of results.
We applied K2 Taxonomer to publicly available HNSCC scRNA-seq data, identifying pertinent tumor cell subtypes, distinguished by cell cycle and epithelial-to-mesenchymal transition, and with analytical projection of this signature onto TCGA-HNSCC bulk RNA-seq data exhibiting association with worse survival in TCGA-HNSCC patients. A transcriptional signature corresponding to suppression of Mtorc1 and Wnt/β-catenin signaling was also identified in a sub-population of HNSCCs, and shown to be associated with improved patient survival. Of notice, highly ranked markers of this signature are significantly associated with gene expression changes altered by E7386 - a novel β-catenin/CBP modulator with an activity profile that closely overlaps with that of ICG-001, but exhibits ~50-100-fold lower EC50 values - suggesting a role for this signaling axis in subsets of HNSCC. In conclusion, taxonomical subtyping with K2 Taxonomer provides a novel framework to expand the scope of applicability of scRNA-seq clustering results, and has revealed potentially novel HNSCC subtypes that offer directions for future studies.
Citation Format: Eric Reed, Takashi Owa, Kenichi Nomoto, Xaralabos Varelas, Maria Kukuruzinska, Stefano Monti. Subtyping of HNSCC single-cell RNA-seq identifies transcriptional programs characterized by suppression of Mtorc1 and Wnt signaling pathways and better patient prognosis [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 4419.