iMIRACLE: an Iterative Multi-View Graph Neural Network to Model Intercellular Gene Regulation from Spatial...

Area of Science:

• Genomics
• Computational Biology
• Systems Biology

Background:

• Spatial transcriptomics enables gene expression analysis within cellular microenvironments.
• Cell-cell communication (CCC) via ligand-receptor (LR) interactions drives cellular responses.
• Existing methods lack robust approaches to link CCC to gene expression changes for biological insights.

Purpose of the Study:

• To develop a novel computational method for modeling intercellular regulation.
• To integrate spatial transcriptomics data with cell-cell communication networks.
• To infer cell-specific gene regulatory scores driven by LR interactions.

Main Methods:

• Developed iMiracle, an iterative multi-view graph neural network.
• Integrated inter- and intra-cellular networks for gene expression estimation.
• Employed iterative learning to address data sparsity in CCC networks.
• Inferred cell-specific ligand-gene regulatory scores using LR pair contributions.

Main Results:

• iMiracle demonstrated superior performance in gene expression imputation across diverse datasets.
• The method outperformed existing approaches in inferring regulatory scores.
• Evaluated on nine simulated and eight real spatial transcriptomics datasets from three platforms.

Conclusions:

• iMiracle effectively models intercellular regulation by integrating gene expression and CCC data.
• The tool enhances the interpretability of spatial transcriptomics by providing cell-specific regulatory insights.
• iMiracle is an open-source software poised to advance the study of cell-cell communication and transcriptional regulation.