Detecting spatially co-expressed gene clusters with functional coherence by graph-regularized convolutional neural...

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Spatial gene expression analysis is crucial for understanding tissue morphology and gene function.
• Conventional clustering methods often overlook spatial co-localization and functional gene relationships.

Purpose of the Study:

• To develop a new method for clustering spatially resolved gene expression data.
• To improve the detection of spatial expression patterns and biological interpretation of gene clusters.

Main Methods:

• A convolutional neural network (CNN) regularized by a protein-protein interaction (PPI) network was developed.
• The method utilizes graph-Laplacian regularization to incorporate gene functional relationships.

Main Results:

• The PPI-regularized CNN successfully detected gene clusters with coherent spatial patterns across 22 Visium datasets.
• The method demonstrated state-of-the-art performance in identifying functionally enriched gene clusters.
• Case studies on mouse kidney and human breast cancer tissues highlighted the detection of spatially co-expressed genes and their morphological context.

Conclusions:

• The PPI-regularized CNN offers an effective approach for analyzing spatial transcriptomics data.
• This method provides valuable biological insights by integrating spatial information and gene functional networks.