Multi-modal domain adaptation for revealing spatial functional landscape from spatially resolved transcriptomics

StereoMM: a graph fusion model for integrating spatial transcriptomic data and pathological images

Spatial omics technologies, generating high-throughput and multimodal data, have necessitated the development of advanced data integration methods to facilitate comprehensive biological and clinical treatment discoveries. Based on the cross-attention concept, we developed an AI learning based toolchain called StereoMM, a graph based fusion model that can incorporate omics data such as gene expression, histological images, and spatial location. StereoMM uses an attention module for omics data interaction and a graph autoencoder to integrate spatial positions and omics data in a self-supervised manner. Applying StereoMM across various cancer types and platforms has demonstrated its robust capability. StereoMM outperforms competitors in identifying spatial regions reflecting tumour progression and shows promise in classifying colorectal cancer patients into deficient mismatch repair and proficient mismatch repair groups. The comprehensive inter-modal integration and efficiency of StereoMM enable researchers to construct spatial views of integrated multimodal features efficiently, advancing thorough tissue and patient characterization.

Spatially Aware Domain Adaptation Enables Cell Type Deconvolution from Multi-Modal Spatially Resolved Transcriptomics

Spatially Resolved Transcriptomics (SRT) offers unprecedented opportunities to elucidate the cellular arrangements within tissues. Nevertheless, the absence of deconvolution methods that simultaneously model multi-modal features has impeded progress in understanding cellular heterogeneity in spatial contexts. To address this issue, SpaDA is developed, a novel spatially aware domain adaptation method that integrates multi-modal data (i.e., transcriptomics, histological images, and spatial locations) from SRT to accurately estimate the spatial distribution of cell types. SpaDA utilizes a self-expressive variational autoencoder, coupled with deep spatial distribution alignment, to learn and align spatial and graph representations from spatial multi-modal SRT data and single-cell RNA sequencing (scRNA-seq) data. This strategy facilitates the transfer of cell type annotation information across these two similarity graphs, thereby enhancing the prediction accuracy of cell type composition. The results demonstrate that SpaDA surpasses existing methods in cell type deconvolution and the identification of cell types and spatial domains across diverse platforms. Moreover, SpaDA excels in identifying spatially colocalized cell types and key marker genes in regions of low-quality measurements, exemplified by high-resolution mouse cerebellum SRT data. In conclusion, SpaDA offers a powerful and flexible framework for the analysis of multi-modal SRT datasets, advancing the understanding of complex biological systems.

Spatially informed graph transformers for spatially resolved transcriptomics

Spatially resolved transcriptomics (SRT) has emerged as a powerful technique for mapping gene expression landscapes within spatial contexts. However, significant challenges persist in effectively integrating gene expression with spatial information to elucidate the heterogeneity of biological tissues. Here, we present a Spatially informed Graph Transformers framework, SpaGT, which leverages both node and edge channels to model spatially aware graph representation for denoising gene expression and identifying spatial domains. Unlike conventional graph neural networks, which rely on static, localized convolutional aggregation, SpaGT employs a structure-reinforced self-attention mechanism that iteratively evolves topological structural information and transcriptional signal representation. By replacing graph convolution with global self-attention, SpaGT enables the integration of both global and spatially localized information, thereby improving the detection of fine-grained spatial domains. We demonstrate that SpaGT achieves superior performance in identifying spatial domains and denoising gene expression data across diverse platforms and species. Furthermore, SpaGT facilitates the discovery of spatially variable genes with significant prognostic potential in cancer tissues. These findings establish SpaGT as a powerful tool for unraveling the complexities of biological tissues.

Spatially aligned graph transfer learning for characterizing spatial regulatory heterogeneity

Spatially resolved transcriptomics (SRT) technologies facilitate the exploration of cell fates or states within tissue microenvironments. Despite these advances, the field has not adequately addressed the regulatory heterogeneity influenced by microenvironmental factors. Here, we propose a novel Spatially Aligned Graph Transfer Learning (SpaGTL), pretrained on a large-scale multi-modal SRT data of about 100 million cells/spots to enable inference of context-specific spatial gene regulatory networks across multiple scales in data-limited settings. As a novel cross-dimensional transfer learning architecture, SpaGTL aligns spatial graph representations across gene-level graph transformers and cell/spot-level manifold-dominated variational autoencoder. This alignment facilitates the exploration of microenvironmental variations in cell types and functional domains from a molecular regulatory perspective, all within a self-supervised framework. We verified SpaGTL's precision, robustness, and speed over existing state-of-the-art algorithms and show SpaGTL's potential that facilitates the discovery of novel regulatory programs that exhibit strong associations with tissue functional regions and cell types. Importantly, SpaGTL could be extended to process multi-slice SRT data and map molecular regulatory landscape associated with three-dimensional spatial-temporal changes during development.