Transformer and graph variational autoencoder to identify microenvironments: A deep learning protocol for spatial transcriptomics

We present transformer and graph variational autoencoder to identify microenvironments (TG-ME), a computational framework that integrates transformer and graph variational autoencoders to dissect spatial niches using spatial transcriptomics and morphological images. This protocol outlines data normalization, spatial transcriptomics integration, morphological feature extraction, and niche profiling. Using deep learning, TG-ME enables robust niche clustering applicable to healthy, tumor, and infected tissues. For complete details on the use and execution of this protocol, please refer to Paniagua et al.