Fig. 3: Single-cell gene expression analysis using HIT-scISOseq.

a Correlation scatter plot showing corresponding UMI counts by cell barcode in NGS (y-axis) and HIT-scISOseq (x-axis) data (Pearson’s correlation coefficient r  = 0.992, n = 1676, p = 0). The NGS and HIT-scISOseq data sets were generated using cDNA of the same corneal limbus sample s1. Dot colors reflect the local density of data points. b Correlation on UMI counts by gene between NGS (y-axis) and HIT-scISOseq (x-axis) data (Pearson’s correlation coefficient r = 0.956, n = 14,513, p = 0). c Correlation of UMI counts by gene between two HIT-scISOseq biological replicate samples (Pearson’s correlation coefficient r = 0.998, n = 13,663, p = 0). d, e UMAP projection of NGS and HIT-scISOseq data. Gene expression profiles were determined independently for each cell cluster using either NGS or HIT-scISOseq. Both the NGS (d) and HIT-scISOseq (e) data sets showed that the four main cell populations could be successfully clustered (differentiated cells, denoted by D, are red; corneal basal cells, denoted by B, are blue; limbal stem cells, denoted by L, are green; and conjunctival cells, denoted by Cj, are purple). The undefined cells: which were potentially immune cells (as indicated in Soure Data), and not used in this analysis as since the main focus of this study is on limbal epithelial cells. f Bar plot showing the percentages of cell barcodes shared between NGS and HIT-scISOseq data in s1 sample. g Heatmap showing the gene expression correlation between NGS and HIT-scISOseq data for each cell cluster in s1 sample. h, i Heatmaps showing the expression of the top15 marker genes of the four major cell clusters in NGS (h) and HIT-scISOseq (i) data sets. The color gradient represents log-transformed and row-normalized counts (each row scaled to a maximum of 1). Upper bars represent cell cluster assignments for individual cells. Source data are provided as a Source Data file.