Fig. 6: Optimized scale-space resampling based on varifold norm for genes and cell type feature spaces of BARseq.

a Shows a single section with 16 selected gene measurements totaling 48 M transcripts. b–d Show the gene with highest probability per particle at 50 μm,100 μm, 200 μm approximations with 82k, 21k and 6k particles, respectively. e–h Show probabilities at each scale of the single Slc17a7 gene. i–l Show a single slice of BARseq cell-segmented data with 104K cells at highest resolution and at 50 μ, 100 μ, and 200 μ approximations with 54k, 18k and 5k cells, respectively. Each cell is colored with the highest probability cell type. m Shows 1 mm × 1.25 mm window of full resolution gene detections for 16 genes (in a), and n shows detections only of Alcam and Nnat. o Shows detections in n overlaid with particles whose estimated feature distributions carry the highest probability of expression for Alcam (yellow) and Nnat (pink) in optimally estimated positions along the curvilinear boundary of tissue.