Transfer learning of multicellular organization via single-cell and spatial transcriptomics

Yecheng Tan,Ai Wang,Zezhou Wang,Wei Lin,Yan Yan,Qing Nie,Jifan Shi

DOI: https://doi.org/10.1101/2024.02.28.582493

2024-06-25

Abstract:Spatial tissues exhibit complex gene expression and multicellular patterns that are difficult to dissect. Single-cell RNA sequencing (scRNA-seq) provides full coverages of genes, but lacking spatial information, whereas spatial transcriptomics (ST) measures spatial locations of individual or group of cells, with more restrictions on gene information. To integrate scRNA-seq and ST data, we introduce a transfer learning method to decipher spatial organization of cells named iSORT. iSORT trains a neural network that maps gene expressions to spatial locations using scRNA-seq data along with ST slices as references. iSORT can find spatial patterns at single-cell scale, identify key genes that drive the patterning, and infer pseudo-growth trajectories using a concept of SpaRNA velocity. Benchmarking on simulation data and comparing with multiple existing tools show iSORT's robustness and accuracy in reconstructing spatial organization. Using our own new human artery datasets, iSORT shows its capability of dissecting atherosclerosis. Applications to a range of biological systems, such as mouse embryo, mouse brain, Drosophila embryo, and human developmental heart, demonstrate that iSORT can utilize both scRNA-seq and ST datasets to uncover multilayer spatial information of single cells.

Bioinformatics

What problem does this paper attempt to address?

The problem this paper attempts to address is: how to analyze the spatial organization patterns of cells within tissues by integrating single-cell transcriptomics (scRNA-seq) and spatial transcriptomics (ST) data. Specifically, the paper proposes a transfer learning method named iSORT, which aims to:

1. **Analyze the spatial organization of cells**: By combining scRNA-seq data with ST data, iSORT can predict the precise location of individual cells within tissues, thereby revealing the spatial distribution patterns of cells.

2. **Identify key genes**: iSORT can identify key genes that play a decisive role in the spatial organization of cells, referred to as Spatial Organization Genes (SOGs). Changes in these genes have a significant impact on the spatial structure of tissues.

3. **Infer pseudo-growth trajectories**: iSORT introduces a new concept—SpaRNA velocity, which projects RNA velocity into physical space to simulate the migration and differentiation trajectories of cells in space.

Through these methods, iSORT not only reconstructs the spatial structure of tissues but also helps researchers understand the dynamic changes of cells within tissues and the role of key genes in this process. The paper validates the effectiveness and robustness of iSORT through multiple benchmark datasets and applications in actual biological systems. For example, the paper demonstrates the application of iSORT in systems such as the human dorsolateral prefrontal cortex (DLPFC), mouse embryos, Drosophila embryos, and human developing hearts. Additionally, the paper explores key genes related to atherosclerosis through experimental data, providing new perspectives for further research into disease mechanisms.