MuSiC2: cell-type deconvolution for multi-condition bulk RNA-seq data.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Cell-type composition in bulk tissues varies and changes during disease.
• Accurate cell-type proportion estimation is crucial for understanding disease pathogenesis.
• Existing deconvolution methods often require matched single-cell RNA sequencing (scRNA-seq) data, which is difficult to obtain for diseased samples.

Purpose of the Study:

• To develop a more accurate method for cell-type deconvolution of bulk RNA sequencing (RNA-seq) data when reference scRNA-seq data is from different clinical conditions.
• To overcome limitations of existing methods that may produce biased estimates due to mismatched sample conditions.

Main Methods:

• Proposed MuSiC2, an iterative estimation procedure extending the original MuSiC method.
• Applied MuSiC2 to bulk RNA-seq data from human pancreatic islets and retina.
• Conducted extensive benchmark evaluations comparing MuSiC2 with traditional MuSiC deconvolution.

Main Results:

• MuSiC2 significantly improved the accuracy of cell-type proportion estimates compared to traditional MuSiC.
• Demonstrated improved performance when bulk RNA-seq samples and scRNA-seq references differed in clinical conditions.
• Successfully applied MuSiC2 to real-world datasets from human pancreatic islets and retina.

Conclusions:

• MuSiC2 offers a more robust and accurate approach for cell-type deconvolution of bulk RNA-seq data, especially when dealing with varying clinical conditions.
• The method facilitates downstream analysis and aids in identifying cellular targets for human diseases.
• MuSiC2 enhances current deconvolution capabilities, providing valuable insights into condition-specific cell-type compositions.