Hidden network preserved in Slide-tags data allows reference-free spatial reconstruction.

Spatial transcriptomics technologies aim to spatially map gene expression in tissues and typically use oligonucleotide array surfaces that have undergone spatial indexing. These arrays are used to capture nucleic acids diffusing from adjacently placed tissues, allowing subsequent sequencing to reveal both gene and position. Slide-tags is a recently developed method by Russell et al. that inverts this principle. Instead of capturing molecules released from the tissue, probes are detached from a pre-decoded bead array and diffused into tissues, tagging nuclei with spatial barcodes. In this work we reanalyze this data and discover a latent, spatially informative cell-bead network formed incidentally from barcode diffusion and the biophysical properties of the tissue. This allows us to treat Slide-tags as a network-based imaging-by-sequencing approach. By optimizing spatial constraints encoded in the cell-bead network structure, we can achieve unassisted tissue reconstruction, a fundamental shift from classical spatial technologies based on pre-indexed arrays.