Identifying Cellular Populations in SCI Spatial Transcriptomics

Sabila Bernard

Mentor: Dr. Robert Suter, Department of Oncology, Lombardi Comprehensive Cancer Center, Georgetown University Medical Center.

Date/Time: December 9th, 2025 at 12:00 PM.

Abstract: Spinal cord injury (SCI) triggers highly localized and rapidly evolving molecular responses, with inflammation, cell loss, and regeneration occurring in distinct spatial domains of the tissue. Because these processes unfold in location-specific patterns, spatial transcriptomics provides a powerful framework for capturing the architecture of SCI pathology.

Spatial transcriptomic profiling is well-suited for studying spinal cord injury (SCI), where transcriptional changes arise in spatially patterned ways and evolve across the injury time course. To support analysis of a Curio Seeker SCI dataset, I implemented a computational workflow focused on downstream preprocessing, spatial aggregation, and spatial feature construction. A Seurat object containing Curio Seeker bead-level data was used as the starting point for subsequent computational analysis. Because individual Curio beads often contain low UMI counts, I developed a custom geometric binning method that overlays a hexagonal spatial grid onto the tissue section and aggregates expression counts for beads falling within each grid cell. This approach increases effective UMI depth, reduces sparsity, and preserves the underlying geometric structure of the tissue.

The resulting binned Seurat objects were then processed with BANKSY, a spatial feature-learning method that combines each bin’s intrinsic expression profile with averaged expression from its nearest spatial neighbors. BANKSY-derived feature matrices supported stable dimensionality reduction, clustering, and the initial identification of spatially organized regions across the SCI time course. This workflow prepares the spatial dataset for later analyses aimed at interpreting cell states, identifying injury-associated patterns, and supporting downstream efforts to characterize inflammatory phenotypes relevant to SCI biology.