Three steps of the scHolography workflow. (1) scHolography takes in ST and SC expression data and ST 2D spatial registration data. Spatial-information components (SICs) are defined for the spatial registration data. ST and SC expression data are integrated. (2) Neural networks are trained with post-integration ST data as input and top SIC values as the target. (3) The trained neural networks are applied to post-integration SC data to predict top SIC values for SC. SIC values are referenced to infer cellcell affinity and construct the stable matching neighbor (SMN) graph. The graph is visualized in 3D.
