Intelligent Spatial Transcriptomics Analysis: From Graph Contrastive Learning to Foundation Models Development

Spatial transcriptomics (ST) has revolutionized our ability to measure gene expression with spatial context, enabling unprecedented insights into tissue architecture and cellular organization.In this talk, we present STAIG (Spatial Transcriptomics Analysis via Image-Aided Graph Contrastive Learning), a computational framework that integrates spatial information with gene expression through graph neural networks and contrastive learning. By incorporating histological images and spatial graphs, STAIG achieves enhanced spatial domain identification across diverse tissue types and ST platforms.Furthermore, we will discuss current challenges in high-resolution spatial transcriptomics, including computational scalability and automated interpretation. We will introduce our ongoing work on developing foundation models for ST analysis, which aims to provide unified, scalable solutions for processing and interpreting high-resolution spatial omics data.