Abstract
While bulk RNA sequencing provides a comprehensive view of transcriptomes, it lacks cell type specificity. Single-cell RNA sequencing (scRNA-seq) overcomes this limitation by providing detailed insights at the individual cell level, though it involves higher costs. Deconvolution methods can estimate cell type proportions in bulk RNA-seq data, but their results may vary based on the scRNA-seq reference data and software used. This study investigates the estimation of muscle fiber type proportions through deconvolution analysis of Longissimus dorsi muscle bulk RNA-seq data from late-gestation Holstein Friesian multiparous cows. Four software tools (i.e., CIBERSORTx, Cellanneal, DeconvR-NNLS, and DeconvR-RLM) were compared using scRNA-seq reference data from Brahman and Wagyu cattle breeds, which included proportions of types I, IIa, and IIx myofibers. Kruskal-Wallis and Dunn's tests revealed that the breed of reference data significantly influenced the proportions of type IIa and IIx muscle fibers across different deconvolution methods. To the best of our knowledge, this is the first study to show that the cattle breed used in reference scRNA-seq data can substantially impact deconvolution outcomes, highlighting a critical consideration for accurate cell type proportion estimation in livestock genomics. These findings suggest that future deconvolution studies should carefully consider breed compatibility between reference and target datasets.