**SpatialGlue: Deciphering Spatial Domains from Spatial Multi-Omics** Yahui Long, Kok Siong Ang, Raman Sethi, Sha Liao, Yang Heng, Lynn van Olst, Shuchen Ye, Chengwei Zhong, Hang Xu, Di Zhang, Immanuel Kwok, Nazihah Husna, Min Jian, Lai Guan Ng, Ao Chen, Nicholas R. J. Gascoigne, David Gate, Rong Fan, Xun Xu & Jimiao Chen Advances in spatial omics technologies have enabled the acquisition of multiple types of data from the same tissue slice. To fully utilize these data, spatially informed methods for data integration are necessary. This paper introduces SpatialGlue, a graph neural network model with a dual-attention mechanism that integrates spatial location and omics measurement within each modality, followed by cross-omics integration. SpatialGlue was tested on various tissue types and technologies, including spatial epigenome-transcriptome and transcriptome-proteome modalities. Compared to other methods, SpatialGlue captured more anatomical details and accurately resolved spatial domains, such as the cortex layers of the brain. It also identified cell types like spleen macrophage subsets located in different zones that were not present in the original data annotations. SpatialGlue scales well with data size and can integrate three modalities. The method combines information from complementary omics modalities to provide a holistic view of cellular and tissue properties. The paper presents a detailed description of the SpatialGlue model, including its architecture and implementation. It demonstrates the effectiveness of SpatialGlue through benchmarking on simulated and experimental spatial multi-omics data, showing superior performance in terms of homogeneity, mutual information, v-measure, AMI, NMI, and ARI metrics. The model's ability to integrate spatial information with individual omics data and across multiple modalities is highlighted, along with its capability to handle different data distributions and feature counts. The results from various datasets, including human lymph node, mouse brain, thymus, and spleen, showcase the model's robustness and broad applicability across different technology platforms. The discussion section emphasizes the importance of spatial information and cross-omics integration in spatial multi-omics data analysis. SpatialGlue's design allows for seamless extension to image-based omics data and future extensions to include images as a modality and integrate multi-omics data from serial tissue sections. The method's computational efficiency and scalability are also noted, making it a valuable tool for analyzing spatial multi-omics data.**SpatialGlue: Deciphering Spatial Domains from Spatial Multi-Omics** Yahui Long, Kok Siong Ang, Raman Sethi, Sha Liao, Yang Heng, Lynn van Olst, Shuchen Ye, Chengwei Zhong, Hang Xu, Di Zhang, Immanuel Kwok, Nazihah Husna, Min Jian, Lai Guan Ng, Ao Chen, Nicholas R. J. Gascoigne, David Gate, Rong Fan, Xun Xu & Jimiao Chen Advances in spatial omics technologies have enabled the acquisition of multiple types of data from the same tissue slice. To fully utilize these data, spatially informed methods for data integration are necessary. This paper introduces SpatialGlue, a graph neural network model with a dual-attention mechanism that integrates spatial location and omics measurement within each modality, followed by cross-omics integration. SpatialGlue was tested on various tissue types and technologies, including spatial epigenome-transcriptome and transcriptome-proteome modalities. Compared to other methods, SpatialGlue captured more anatomical details and accurately resolved spatial domains, such as the cortex layers of the brain. It also identified cell types like spleen macrophage subsets located in different zones that were not present in the original data annotations. SpatialGlue scales well with data size and can integrate three modalities. The method combines information from complementary omics modalities to provide a holistic view of cellular and tissue properties. The paper presents a detailed description of the SpatialGlue model, including its architecture and implementation. It demonstrates the effectiveness of SpatialGlue through benchmarking on simulated and experimental spatial multi-omics data, showing superior performance in terms of homogeneity, mutual information, v-measure, AMI, NMI, and ARI metrics. The model's ability to integrate spatial information with individual omics data and across multiple modalities is highlighted, along with its capability to handle different data distributions and feature counts. The results from various datasets, including human lymph node, mouse brain, thymus, and spleen, showcase the model's robustness and broad applicability across different technology platforms. The discussion section emphasizes the importance of spatial information and cross-omics integration in spatial multi-omics data analysis. SpatialGlue's design allows for seamless extension to image-based omics data and future extensions to include images as a modality and integrate multi-omics data from serial tissue sections. The method's computational efficiency and scalability are also noted, making it a valuable tool for analyzing spatial multi-omics data.