Comprehensive evaluation of transcriptome-based cell-type quantification methods for immuno-oncology.

Area of Science:

• Immunology
• Computational Biology
• Bioinformatics

Background:

• Immune cell composition in the tumor microenvironment (TME) impacts cancer progression and therapy outcomes.
• Bulk RNA-sequencing (RNA-seq) is often used to infer immune cell types when single-cell methods are unavailable.
• A systematic evaluation of computational deconvolution tools for immune cell estimation from RNA-seq data is lacking.

Purpose of the Study:

• To systematically benchmark computational methods for estimating immune and stromal cell composition from bulk RNA-seq data.
• To provide a comprehensive evaluation of deconvolution tool accuracy across various cell types and datasets.
• To offer guidelines for improving cell type signature definitions and deconvolution strategies.

Main Methods:

• Developed a systematic benchmarking approach using simulated bulk RNA-seq samples derived from a single-cell RNA-seq dataset of the TME.
• Assessed the accuracy of seven computational deconvolution methods for estimating nine different immune and stromal cell types.
• Validated performance using independent, publicly available gold-standard estimates and analyzed over 1800 samples.

Main Results:

• Demonstrated high accuracy of computational deconvolution for well-defined cell-type signatures.
• Identified areas for improvement in handling fuzzy cell-type signatures.
• Provided an exhaustive evaluation across multiple methods, cell types, and datasets.

Conclusions:

• Computational deconvolution is a valuable tool for estimating immune cell composition from bulk RNA-seq.
• Refining cell population definitions and identifying reliable signatures are crucial for future advancements.
• The study provides a reproducible benchmark pipeline and an R package for integrated deconvolution.