Fig. 1: Overview of the computational framework of SONAR.

SONAR is designed for cell-type deconvolution of spatial transcriptomic data. a SONAR inputs spatial transcriptomics expression matrix (top) with coordinates information (middle) and an annotated scRNA-seq data as cell type reference (bottom). b The Poisson-Gamma regression model of SONAR. c The spatial (kernel) weight and the adaptive tuning strategy (pre-clustering, and elastic weighting) applied in SONAR. d SONAR outputs a spatial map of cell types and the detailed cell type composition of each spot.