Cell-Type Deconvolution of Equine BALF RNA-Seq: A Critical Comparison with Matched Single-Cell Data

Area of Science:

• Veterinary Genomics
• Computational Biology
• Immunology

Background:

• Bulk RNA sequencing (RNA-seq) averages gene expression across cell types.
• Computational deconvolution methods aim to infer cell composition and gene expression from bulk data.
• The performance of deconvolution in equine samples, particularly for severe equine asthma (SEA), is unevaluated.

Purpose of the Study:

• Assess computational deconvolution's ability to recover cellular composition in equine bronchoalveolar lavage fluid (BALF).
• Evaluate deconvolution's capacity to identify cell type-specific differential gene expression in SEA.
• Compare deconvolution performance against single-cell RNA sequencing (scRNA-seq) as a reference.

Main Methods:

• Generated bulk RNA-seq data from BALF of SEA and control horses.
• Utilized matched scRNA-seq data as a reference for deconvolution.
• Evaluated deconvolution accuracy by comparing estimates with scRNA-seq cell proportions and differential expression analysis.

Main Results:

• Deconvolution captured mRNA-derived proportions, with moderate agreement to cell counts (r=0.62), improving with mRNA correction (r=0.83).
• Near-perfect concordance (r=0.98) was observed with mRNA-weighted scRNA-seq proportions.
• Cell type-specific performance varied, and differential expression recovery was inconsistent, showing signal spillover.

Conclusions:

• Deconvolution of bulk RNA-seq did not reliably estimate cell counts in equine BALF.
• Deconvolution provided limited biological insight beyond conventional bulk RNA-seq analysis.
• Single-cell RNA sequencing is crucial for resolving cell type-specific disease mechanisms in equine asthma.