CDSeqR: fast complete deconvolution for gene expression data from bulk tissues.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Biological tissues are composed of diverse cell populations, making bulk tissue gene expression analysis complex.
• Understanding cell-type-specific contributions is crucial for interpreting overall tissue gene expression patterns.
• Existing bulk RNA-seq data analysis often overlooks crucial tissue heterogeneity.

Purpose of the Study:

• To present CDSeqR, an R implementation of the CDSeq computational method for analyzing bulk RNA sequencing data.
• To improve the computational efficiency and add new functionalities for cell type annotation.
• To enable the recovery of cell-type-specific expression information from bulk RNA-seq data.

Main Methods:

• Developed a novel strategy to enhance computational efficiency in speed and memory usage for the CDSeq method.
• Implemented a new function within CDSeqR for annotating estimated cell types using single-cell RNA sequencing (scRNA-seq) data and marker genes.
• Validated CDSeqR using synthetic data, real cell mixtures, and bulk RNA-seq datasets from TCGA and GTEx projects.

Main Results:

• Achieved significant performance improvements in speed and memory usage compared to the original MATLAB implementation.
• Enabled straightforward interpretation and visualization of CDSeq-estimated cell types through the new annotation function.
• Demonstrated the utility of CDSeqR on diverse datasets, including large-scale cancer and tissue expression data.

Conclusions:

• CDSeqR provides a powerful tool for the in silico dissection of bulk RNA-seq data, addressing limitations of bulk analyses.
• The package facilitates the recovery of cell-type-specific expression profiles, offering deeper insights into tissue microenvironments.
• CDSeqR enhances the utility of existing bulk RNA-seq repositories like TCGA and GTEx for disease and cell-cell interaction studies.