Abstract PR004: A comprehensive pan-sarcoma single-cell transcriptomic meta-analysis reveals shared molecular programs across subtypes
TL;DR
A systematic meta-analysis of all publicly available single cell RNA sequencing datasets of all human sarcomas reveals shared molecular programs across subtypes, establishing a foundational resource for identifying conserved transcriptional programs across sarcomas, with implications for future mechanistic studies and therapeutic repurposing for these rare cancers.
Abstract
Sarcomas are rare malignancies that encompass diverse histologic subtypes, despite their shared mesenchymal origins. Given their rarity, completing large-scale studies that aim to elucidate sarcoma biology and derive meaningful therapeutic advancements remains challenging. Identifying shared and distinct molecular programs across sarcomas presents a unique opportunity to efficiently repurpose and advance the therapeutic management of these rare tumors. The purpose of this study was to define the transcriptomic landscape of all human sarcomas. To do this, we performed a systematic meta-analysis of all publicly available single cell RNA sequencing (scRNA-seq) datasets of all human sarcomas. Raw datasets were methodically compiled from the Gene Expression Omnibus (GEO) database. After applying standard quality control metrics, the scanpy pipeline was applied and all datasets were integrated with scVI. Cell types were annotated using a combination of canonical markers and by inferring copy number variation using infer CNV. We screened 1,267 scRNA-seq datasets from 991 studies and identified 210 samples from 34 datasets that met all inclusion and exclusion criteria. After quality control, these datasets encompassed 15 different types of sarcomas, comprised of over a million cells. Following integration and annotation, these tumors were found to have heterogenous tumor microenvironments comprised of several cell types, including tumor, immune, endothelial, and fibroblast lineages, with varying compositions across sarcoma subtypes. CNV inference further distinguished tumor cells from normal cell populations. Among the tumor cells, there were several transcriptional programs shared across sarcomas, including processes related to migration and invasion (PARD3+/AUTS2+/AGAP1+ cells), high translational activity (NPM1+/B2M+/RPL24+ cells), and mesenchyme-like phenotype with matrix remodeling features (THBS2+/COL1A1+/COL6A3+ cells). In summary, we present a comprehensive meta-analysis of all human sarcoma single-cell transcriptomic datasets ever published. To our knowledge, this represents the largest integrated analysis of human sarcomas performed to date. Future analyses will include correlation of transcriptomic signatures with patient outcomes using bulk RNA sequencing data through The Cancer Genome Atlas, along with deriving targeted drug predictions using the drug2cell pipeline. This study establishes a foundational resource for identifying conserved transcriptional programs across sarcomas, with implications for future mechanistic studies and therapeutic repurposing for these rare cancers. Maria Korah, James Agolia, Renceh AB. Flojo, Biren Reddy, Kaylin Yip, Deshka Foster, Michael Longaker, Daniel Delitto. A comprehensive pan-sarcoma single-cell transcriptomic meta-analysis reveals shared molecular programs across subtypes [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Breaking Barriers in the Fight against Rare Cancers; 2026 Jul 18-20; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2026;86(14_Suppl):Abstract nr PR004.