Joint dimension reduction and clustering analysis of single-cell RNA-seq and spatial transcriptomics data

Area of Science:

• Computational biology
• Bioinformatics
• Spatial transcriptomics

Background:

• Sequential dimension reduction and clustering may yield suboptimal results.
• Low-dimensional embeddings might not align with inferred cluster labels.
• Existing methods lack integrated analysis for spatial transcriptomics data.

Purpose of the Study:

• To develop a novel computational method for simultaneous dimension reduction and spatial clustering.
• To improve the accuracy and biological relevance of feature extraction in spatial transcriptomics.
• To provide a unified framework for analyzing tissue organization and spatial relationships.

Main Methods:

• Developed Dimension-Reduction Spatial-Clustering (DR-SC), a unified framework.
• Integrated dimension reduction and spatial clustering using a latent hidden Markov random field model.
• Employed an expectation-maximization algorithm with iterative conditional mode for efficient computation.

Main Results:

• DR-SC achieved accurate spatial clustering and effective extraction of biologically informative features.
• The method demonstrated superior performance compared to existing clustering and spatial clustering approaches.
• DR-SC improved downstream trajectory inference and visualization.

Conclusions:

• DR-SC offers a powerful, integrated approach for spatial transcriptomics analysis.
• The method enhances the understanding of tissue spatial organization and cellular heterogeneity.
• DR-SC provides a scalable and data-driven solution for complex biological data analysis.