Cross-Platform Comparative Spatial Transcriptomics in High-Grade Serous Carcinoma.

PURPOSE: High-grade serous carcinoma of the ovary/fallopian tube (HGSC) is a remarkably heterogeneous tumor. The purpose of this study was to directly compare the reproducibility of transcriptomic profiles and biologically relevant molecular features across 2 spatial transcriptomics platforms-GeoMx Digital Spatial Profiler (GeoMx) and Visium Spatial Gene Expression (Visium)-using artificial intelligence (AI)-defined tumor regions associated with clinical outcome.

MATERIALS AND

METHODS

A combination of AI-identified tumor regions and Visium Spatial Gene Expression (Visium) was previously applied to formalin-fixed, paraffin-embedded HGSC samples to identify molecular features associated with disease outcome. Here, we apply GeoMx Digital Spatial Profiler (GeoMx) to the same formalin-fixed, paraffin-embedded cohort (n = 16) to directly compare transcriptional profiles between platforms. We analyzed AI-identified regions, including high-confidence (HC) and lower-confidence "background tumor" areas from the same tumor.

RESULTS

A total of 8011 differentially expressed genes were identified by both platforms, showing strong concordance (r = 0.72, P = 2.2 × 10-16), and pathway enrichment patterns, including tumor necrosis factor-α signaling and p53 pathways, were similar. Outcome-related genes of interest exhibited distinct expression patterns between HC and background tumor regions: JUN, ITGB8, KRT7, and TACSTD2 were enriched in the short platinum-free interval group, whereas PPIB, GSTP1, and LPAR3 were associated with long platinum-free interval. Additionally, RNA in situ hybridization validated GeoMx-identified HC region-specific markers. Higher NR1D2 and TREX1 expression in short platinum-free interval tumors and elevated H1-5 and HMGB3 expression in long platinum-free interval tumors were detected, consistent with the differential expression trends observed in GeoMx.

CONCLUSIONS

Together, our results demonstrate that key transcriptomic features of HGSC are consistently captured across spatial transcriptomic technologies, supporting their translational utility. The ability to accurately replicate transcriptomic profiles across 2 different platforms expands translational applications and enhances the identification of biologically significant molecular features relevant to HGSC outcomes.