#5 · Single-cell

Spatial and single-cell models infer multicellular motifs, regulators and disease links

The common thread is inference that makes explicit use of spatial or single-cell organization rather than bulk averages. Spatial methods address segmentation, motif discovery and signal purity: SpatialQuery recovers recurrent multicellular motifs, SpaCNA aggregates neighbor information to detect copy number alterations, TENGU segments Visium HD data, and the Xenium study characterizes noise and purifies intracellular signal with SPLIT. Single-cell methods link cell-level data to disease and regulation, including SCADS, which combines GWAS summary statistics with scATAC-seq, and a pancreatic islet digital twin that validates STAT1 and ZEB1 regulators in silico. Other items address multiplet detection, isoform-aware classification and Hi-C dissimilarity. The set is heterogeneous; the shared element is the structure of the inferential model itself, not a single biological question.