SCDC: bulk gene expression deconvolution by multiple single-cell RNA sequencing references.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Bulk RNA sequencing (RNA-seq) averages gene expression, masking cell-type specific information.
• Single-cell RNA sequencing (scRNA-seq) provides high-resolution transcriptomic data but is often limited in sample throughput.
• Deconvolution methods aim to infer cell-type proportions from bulk RNA-seq data using reference profiles.

Purpose of the Study:

• To develop and validate SCDC, a novel deconvolution method for bulk RNA-seq data.
• To leverage multiple scRNA-seq reference datasets to improve deconvolution accuracy.
• To address batch effects inherent in integrating data from diverse scRNA-seq sources.

Main Methods:

• SCDC utilizes cell-type specific gene expression profiles from multiple scRNA-seq datasets.
• An ENSEMBLE approach integrates deconvolution results from different scRNA-seq references.
• Performance is benchmarked against existing methods using in silico and experimental mixed cell line data.

Main Results:

• SCDC demonstrates superior accuracy in cell-type decomposition compared to existing methods.
• The ENSEMBLE framework effectively integrates heterogeneous scRNA-seq data, mitigating batch effects.
• Application to human pancreatic islets and mouse mammary glands shows improved consistency and phenotype associations.

Conclusions:

• SCDC offers a robust and accurate solution for cell-type deconvolution from bulk RNA-seq.
• The ENSEMBLE strategy enhances the reliability of deconvolution by integrating diverse scRNA-seq references.
• SCDC facilitates deeper biological insights by accurately linking cell-type proportions to phenotypic data.