Spatial Spot Deconvolution: Unmixing Tissue Niches (2D)

Every spatial transcriptomics spot pools mRNA from several neighbouring cells at once, so the raw readout is a blend of cell types rather than a clean single-cell signal. This 2D simulator lays a grid of spots over a tissue you can paint yourself — pick a cell type and drag across the grid to move that niche's true anchor — mixes each spot's observed gene signal from a fixed reference signature matrix plus adjustable noise, then runs the same non-negative multiplicative-update deconvolution used by real tools like Cell2location to unmix each spot back into per-cell-type proportions. Stacked tiles at every spot show the algorithm's live estimate converging toward the ground truth as you increase the iteration count, while sliders let you dial in how mixed the tissue is and how noisy the measurement is.