Fig. 1

Overview of SEDR. SEDR learns a low-dimensional latent representation of gene expression embedded with spatial information by jointly training a masked self-supervised deep autoencoder and a variational graph convolutional autoencoder. The low-dimensional embedding produced by SEDR can be used for downstream visualization, spot clustering, trajectory inference, and batch effect correction. The reconstructed feature matrix can be used to impute the raw gene expression with dropouts