空間的トランスクリプトミクスの解像度を高める解釈可能なマルチモーダル次元削減フレームワークSpaHDmap

Spatial transcriptomics (ST) technologies revolutionized tissue architecture studies by capturing gene expression with spatial context. However, high-dimensional ST data often have limited spatial resolution and exhibit considerable noise and sparsity, posing substantial challenges in deciphering subtle spatial structures and underlying biological activities. Here we introduce 'spatial high-definition embedding mapping' (SpaHDmap), an interpretable dimension reduction framework that enhances spatial resolution by integrating ST gene expression with high-resolution histology images. SpaHDmap incorporates non-negative matrix factorization into a deep learning framework, enabling the identification of high-resolution spatial metagenes (embeddings). Furthermore, SpaHDmap can simultaneously analyse multiple samples and is compatible with various types of histology images. Extensive evaluations on synthetic, public and newly sequenced ST datasets from various technologies and tissue types demonstrate that SpaHDmap can effectively produce high-resolution spatial metagenes, and detect refined spatial structures. SpaHDmap represents a powerful approach for integrating ST data and histology images, offering deeper insights into complex tissue structures and functions.