Poster #19: Lixia Chen Wu MS

STCS: A Platform-Agnostic Framework for Cell-Level Reconstruction in Sequencing-Based Spatial Transcriptomics

Lixia Chen Wu, MS¹˒²˒³, Xinyu Hu, MS¹˒²˒⁴, Fengwei Zhan, BS¹˒²˒⁵, Chuhanwen Sun, MS¹˒²˒⁴, Jose Gonzales, BS¹˒²˒⁶, Rachel Ofer, BA¹˒²˒⁴, Tyler Tran, PhD¹˒², Michael P. Verzi, PhD¹˒², Liping Liu, PhD⁷˒⁸, and Jiekun Yang, PhD¹˒² . ¹Department of Genetics, Rutgers University, Piscataway, NJ, USA ²Human Genetics Institute of New Jersey, Rutgers University, Piscataway, NJ, USA ³Department of Computer Science, Columbia University, New York, NY, USA ⁴Graduate Program of Molecular Biosciences, Rutgers University, Piscataway, NJ, USA ⁵Graduate Program of Quantitative Biomedicine, Rutgers University, Piscataway, NJ, USA ⁶Graduate Program of Biomedical Sciences, Rutgers University, Piscataway, NJ, USA ⁷Department of Mathematics, Rutgers University, Piscataway, NJ, USA ⁸Department of Mechanical and Aerospace Engineering, Rutgers University, Piscataway, NJ, USA

Sequencing-based spatial transcriptomics platforms such as Visium HD and Stereo-seq achieve transcriptome-wide coverage at subcellular resolution, yet their measurements are defined over spatially barcoded units rather than biologically segmented cells. Reconstructing coherent cell-level expression profiles from these data remains a central computational challenge. Here, we introduce Spatial Transcriptomics Cell Segmentation (STCS), a platform-agnostic framework that reconstructs single-cell expression profiles by assigning spatial units to nuclei segmented from paired H&E images, using a combined transcriptomic and spatial distance. STCS is governed by two interpretable parameters that can be selected using reference-free internal metrics. We evaluated STCS in Visium HD human lung cancer, mouse jejunum samples spanning multiple radiation injury and regeneration time points, and Stereo-seq mouse brain tissues. In lung cancer, comparison with matched Xenium data demonstrated accurate spatial reconstruction, while STCS achieved a broad-lineage identity margin of 0.195, approaching the Xenium reference (0.204) and substantially exceeding Bin2Cell, a morphology-based cell reconstruction method (0.023), and the standard 10x Space Ranger nucleus-based cell segmentation algorithm (0.022). STCS also yielded the highest pseudobulk similarity to a single-cell RNA-seq reference. In mouse jejunum, STCS improved agreement with reference profiles, with higher pooled per-cell cosine similarity than 10x (0.824 vs. 0.814). In the mouse brain, evaluation against an unmatched reference tissue section demonstrated accurate spatial organization, with STCS outperforming STP, a cell reconstruction method developed and evaluated on Stereo-seq data, and the nuclei-only baseline in spatial coherence, transcriptomic concordance, and recovery of cell-type composition. Together, these results establish STCS as a robust, fully open-source framework for single-cell reconstruction across sequencing-based spatial transcriptomics technologies.