981-SpaHDmap: for Multimodal Spatial Transcriptomics

SpaHDmap is an innovative multimodal framework designed to enhance the resolution and interpretability of spatial transcriptomics (ST) data. By integrating gene expression profiles with high-resolution histology images, the tool overcomes common challenges like data sparsity and noise to reveal subtle tissue structures. It utilizes a deep learning architecture combined with non-negative matrix factorization to produce "spatial metagenes" that map biological activities at a near-pixel level. Evaluations across synthetic and real-world datasets, including cancer and brain tissues, prove its superior accuracy in identifying complex spatial domains. Furthermore, SpaHDmap supports simultaneous analysis of multiple samples and various imaging types, providing deeper insights into cellular microenvironments. This technical report highlights its potential to revolutionize how researchers visualize and understand the functional organization of complex biological systems.References:Tang J, Chen Z, Qian K, et al. The interpretable multimodal dimension reduction framework SpaHDmap enhances resolution in spatial transcriptomics[J]. Nature Cell Biology, 2026: 1-15.前往小宇宙评论区与主播互动