<i>t</i>-SNE projection of spots and a violin plot.

Principle component analysis (PCA) was run on the normalized filtered feature-barcoded matrix to reduce the number of feature dimensions prior to clustering. For gene expression after PCA, t-distributed neighbor embedding (t-SNE) was run to visualize spots in a 2-D space, followed by clustering to group spots with similar expression profiles. t-SNE colored by UMI counts per spot (A) and spots by clustering (B) are shown for the entire data set. Comparative expression (C) analysis revealed variability in the spatiotemporal distribution of HIF-1a expression across 16 clusters for the entire data set.