Recovering Spatially-Varying Cell-Specific Gene Co-expression Networks for Single-Cell Spatial Expression Data

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell technologies offer unprecedented spatial expression profiling.
• Gene co-expression networks reveal gene-gene dependencies and aid functional annotation.
• Existing methods may not fully leverage spatial information for cell-specific networks.

Purpose of the Study:

• To develop a computationally efficient algorithm for reconstructing spatially-varying, cell-specific gene co-expression networks.
• To improve the accuracy of gene network estimation in single-cell spatial expression data.
• To enable prediction of gene co-expression networks in unobserved tissue regions.

Main Methods:

• A two-step algorithm is proposed: 1. Estimate cell-type-specific gene expression covariance matrices. 2. Construct cell-specific networks using spatial locations and neighboring cell information.
• The method employs Bayesian posterior estimates with empirical priors derived from neighboring cells.
• The algorithm can interpolate gene co-expression networks for uncaptured tissue positions.

Main Results:

• The proposed algorithm demonstrates superior estimation accuracy compared to traditional cell-type-specific and non-spatial cell-specific network methods.
• Simulations confirm the advantages of incorporating spatial information.
• Application to real-world datasets yielded biologically meaningful findings.

Conclusions:

• The developed algorithm effectively recovers spatially-varying cell-specific gene co-expression networks.
• This approach enhances the understanding of spatial genomic features at the cell level.
• The method provides a valuable tool for analyzing single-cell spatial expression data and discovering novel gene interactions.