China National Center for Bioinformation, Beijing, 100101, China.
Beijing Institute of Genomics, Chinese Academy of Sciences, Beijing, 100101, China.
Genome Biol. 2024 Oct 22;25(1):278. doi: 10.1186/s13059-024-03421-5.
Spatial transcriptomics technologies have been widely applied to decode cellular distribution by resolving gene expression profiles in tissue. However, sequencing techniques still limit the ability to create a fine-resolved spatial cell-type map. To this end, we develop a novel deep-learning-based approach, STASCAN, to predict the spatial cellular distribution of captured or uncharted areas where only histology images are available by cell feature learning integrating gene expression profiles and histology images. STASCAN is successfully applied across diverse datasets from different spatial transcriptomics technologies and displays significant advantages in deciphering higher-resolution cellular distribution and resolving enhanced organizational structures.
空间转录组学技术已被广泛应用于通过解析组织中的基因表达谱来解码细胞分布。然而，测序技术仍然限制了创建精细分辨率空间细胞图谱的能力。为此，我们开发了一种新的基于深度学习的方法 STASCAN，通过整合基因表达谱和组织学图像的细胞特征学习，来预测仅具有组织学图像的捕获或未知区域的空间细胞分布。STASCAN 成功地应用于来自不同空间转录组学技术的不同数据集，并在解析更高分辨率的细胞分布和解决增强的组织结构方面显示出显著的优势。
