PAST: latent feature extraction with a Prior-based self-Attention framework for Spatial Transcriptomics

Abstract

Rapid advances in spatial transcriptomics (ST) have revolutionized the interrogation of spatial heterogeneity and increased the demand for comprehensive methods to effectively characterize spatial domains. As a prerequisite for ST data analysis, spatial domain characterization is a crucial step for downstream analyses and biological implications. Here we propose PAST, a variational graph convolutional auto-encoder for ST, which effectively integrates prior information via a Bayesian neural network, captures spatial patterns via a self-attention mechanism, and enables scalable application via a ripple walk sampler strategy. Through comprehensive experiments on datasets generated by different technologies, we demonstrated that PAST could effectively characterize spatial domains and facilitate various downstream analyses, including ST visualization, spatial trajectory inference and pseudo-time analysis, by integrating spatial information and reference from various sources. Besides, we also show the advantages of PAST for accurate annotation of spatial domains in newly sequenced ST data and biological implications in the annotated domains.