Gene set optimization for single cell transcriptomics.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) offers deep biological insights but presents analytical challenges like low statistical power and complex data characteristics.
• Existing gene set collections, designed for bulk tissue analysis, are not optimal for scRNA-seq due to inherent data differences, including noise and sparsity.
• Gene set testing (pathway analysis) is a promising approach to address these challenges, but requires adaptation for single-cell data.

Purpose of the Study:

• To develop a procedure for customizing existing gene set collections for cell type-specific analysis of scRNA-seq data.
• To enhance the statistical power and interpretability of gene set testing in scRNA-seq studies.
• To leverage human cell type-specific gene expression data for improved pathway analysis.

Main Methods:

• Customization of gene set collections by computing cell type-specific gene and gene set weights.
• Utilizing mean gene expression data from the Human Protein Atlas (HPA) Single Cell Type Atlas, profiling 81 human cell types.
• Application of cell type-specific weights to filter or adjust gene set collections for scRNA-seq analysis.

Main Results:

• Demonstrated significant improvements in gene set testing power and interpretability.
• Successfully applied the method to analyze immune cell scRNA-seq data using gene sets from the Molecular Signatures Database (MSigDB).
• Showcased the effectiveness of accounting for cell type-specificity in pathway analysis.

Conclusions:

• Customizing gene set collections based on cell type-specific expression patterns is crucial for effective scRNA-seq analysis.
• The developed procedure enhances the utility of pathway analysis for single-cell transcriptomic data.
• This approach offers a powerful tool for deeper biological interpretation of complex single-cell experiments.