Cell Simulation as Cell Segmentation

Spatial transcriptomics has grown rapidly in scale and adoption, but the quality and interpretability of the data is marred by the limited ability of prior cell segmentation methods to accurately assign transcripts to cells. Segmentation error causes systematic misidentification of cell types and profoundly confounds many spatial analyses. To help address this issue, we developed a probabilistic segmentation algorithm, Proseg, which uses the spatial distribution of transcripts to determine plausibly boundaries, dramatically reducing the degree of spurious coexpression downstream false positive results. Recent updates to the method adds support for Visium HD, handling both imaging- and barcode-based transcriptomic data with the same unified segmentation framework.