Characterizing High-Grade Serous Ovarian Carcinoma Heterogeneity via Spatial Transcriptomics Deconvolution and Tumor Signature Analysis

Epithelial Ovarian Cancer is recognized as the seventh leading cause of cancer-related mortality among women worldwide. Among its subtypes, High-Grade Serous Ovarian Carcinoma (HGSOC) is characterized by pronounced intratumoral heterogeneity, which complicates treatment strategies and hinders accurate prognostic assessment. In this study, all publicly available Spatial Transcriptomics datasets generated using the 10x Genomics Visium platform, originating from four independent studies and hosted on GEO and Zenodo, were included. The dataset comprises 19 HGSOC samples collected following interval debulking surgery which is considered the first-line therapy for this subtype of tumour. Second-line therapy is proposed for patients with unresectable intraperitoneal dissemination and consists of neoadjuvant chemotherapy followed by surgery; this cohort includes 8 of the 19 patients. Each sample was individually processed and filtered, with low-quality spots removed and genes with high mitochondrial or ribosomal expression excluded. The reference-based algorithm CARD was subsequently employed to estimate cell populations within the tissue and annotate regions based on their predominant cell-type composition, distinguishing tumor-rich areas from those dominated by immune or stromal components. Additionally, the signifinder Bioconductor package was used to calculate the expression of established tumor-related gene signatures curated from the literature. The integration of deconvolution and signature analysis provided a comprehensive and nuanced characterization of the tumor microenvironment, revealing distinct molecular processes that underpin HGSOC. Overall, this study establishes a preliminary framework for the spatially resolved analysis of tumor heterogeneity in HGSOC.