Effective methods for bulk RNA-seq deconvolution using scnRNA-seq transcriptomes.

Area of Science:

• Genomics
• Computational Biology
• Biotechnology

Background:

• Single-cell RNA sequencing (scRNA-seq) offers detailed tissue composition analysis but is costly and operationally demanding.
• Computational deconvolution methods aim to infer cell composition from bulk samples using scRNA-seq data, but their efficacy is debated.

Purpose of the Study:

• To systematically evaluate computational deconvolution methods for single-cell RNA sequencing (scRNA-seq) data.
• To identify biases in scRNA-seq assays and data preprocessing.
• To develop an improved method for accurate cell-type abundance estimation.

Main Methods:

• Conducted a systematic evaluation of deconvolution methods using datasets with known or scRNA-seq-estimated compositions.
• Analyzed biases common to 10X Genomics scRNA-seq assays.
• Developed and applied the Single-cell RNA Quantity Informed Deconvolution (SQUID) method, integrating RNA-seq transformation and dampened weighted least-squares deconvolution.

Main Results:

• Identified common biases in scRNA-seq assays and highlighted the importance of data preprocessing and method selection.
• Demonstrated that concurrent RNA-seq and scRNA-seq profiles enhance preprocessing and deconvolution accuracy.
• SQUID consistently outperformed other methods in predicting cell mixture and tissue sample compositions.

Conclusions:

• Concurrent analysis of RNA-seq and scRNA-seq profiles with SQUID yields accurate cell-type abundance estimates.
• This accuracy improvement is crucial for identifying cancer cell subclones that predict outcomes in pediatric cancers.
• Enhanced deconvolution accuracy is vital for advancing applications in the life sciences.