High resolution spatial trancrioptomics combined with deep learning-based image segmentation enables single-cell spatial profiling of archival FFPE kidney tissue from patients with idiopathic nephrotic syndrome

We have perfomed Visium HD spatial transcriptomics on archival FFPE kidney biopsies from patients with different manifistations of nephrotic syndrome, i.e., steriod sensitive, frequent relapsing and/or steroid dependent, and focal segmental glomerulosclerosis. Using a pre-trained deep learning model from StarDist to segment nuclei, we reconstructed single-nucleus level spatial trancriptomic datasets through custom binning of subcellular capture areas. We integrated the datasets from the different patients and indentified conserved cell types. Moreover, we identified patient specific transcriptomic differences in the podocyte cell population. This article demonstrates the usefullness of this approach to future studies of pathophysilogical mechanisms in idiopathic nephrotic syndrome.