Fig. 1: Interpretable deep generative model for spatial transcriptomics analysis.
From: Interpretable spatially aware dimension reduction of spatial transcriptomics with STAMP
a, STAMP workflow. STAMP takes in the gene expression values and a neighborhood graph built from the spot locations. STAMP covers three different analysis scenarios, single sample, multi-sample and time-series modules. STAMP uses black-box variational inference and simplified graph convolutions network (SGCN) to find the variational parameters. b, Capabilities of STAMP. The gene module scores wkg and topic proportions znk are outputs of STAMP. These outputs can be used for downstream analysis such as horizontal, vertical integration and time-series analysis.
