SpaCell: integrating tissue morphology and spatial gene expression to predict disease cells

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Spatial transcriptomics (ST) technology measures gene expression within intact tissues, correlating it with morphology.
• Current ST analysis methods often neglect image pixel data, limiting the integration of gene expression and tissue structure.

Purpose of the Study:

• To develop a novel deep learning software, SpaCell, for integrating spatial transcriptomics and tissue imaging data.
• To leverage pixel intensity information alongside gene expression for enhanced biological insights.

Main Methods:

• Developed SpaCell, a user-friendly deep learning software.
• Integrated millions of pixel intensity values with thousands of gene expression measurements from spatially barcoded spots.
• Utilized deep learning models for cell type identification and tissue image analysis.

Main Results:

• The integrated approach significantly outperformed methods using only gene-count or imaging data.
• SpaCell achieved high resolution and accuracy in identifying cell types and predicting tissue image labels.
• Demonstrated the quantitative link between gene expression and tissue morphology through integrated analysis.

Conclusions:

• SpaCell offers a powerful new tool for analyzing spatial transcriptomics data by incorporating imaging information.
• The software enhances the accuracy and resolution of cell type identification and tissue image analysis.
• This integrated approach provides deeper insights into the relationship between gene expression and tissue architecture.