Adding highly variable genes to spatially variable genes can improve cell type clustering performance in spatial...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Spatial transcriptomics enables transcriptome analysis within tissue context.
• Spatially variable (SV) genes exhibit spatial autocorrelation and are used for clustering.
• Highly variable (HV) genes show significant cell-to-cell expression variation and are conventionally used for clustering.

Purpose of the Study:

• To evaluate if incorporating highly variable (HV) genes alongside spatially variable (SV) genes improves cell type clustering in spatial transcriptomics data.
• To compare the performance of HV genes, SV genes, and their combined set for clustering.

Main Methods:

• Tested clustering performance using HV genes, SV genes, and their union (concatenation).
• Utilized over 50 diverse spatial transcriptomics datasets across multiple platforms.
• Employed a range of spatial and non-spatial metrics for evaluation.

Main Results:

• Combining HV genes and SV genes demonstrated improved overall cell-type clustering performance.
• The integrated gene set outperformed individual gene sets in various datasets and metrics.

Conclusions:

• The union of highly variable and spatially variable genes offers a more robust approach for cell type identification in spatial transcriptomics.
• This combined strategy enhances the accuracy and reliability of spatial transcriptomics analyses.