It has been shown that tail fat content varies significantly among sheep breeds and plays a significant role in meat quality. Recently, significant efforts have been made to understand the physiological, biochemical, and genomic regulation of fat deposition in sheep tails in order to unravel the mechanisms underlying energy storage and adipose tissue lipid metabolism. RNA-seq has enabled us to provide a high-resolution snapshot of differential gene expression between fat- and thin-tailed sheep breeds. Therefore, three RNA-seq datasets were meta-analyzed for the current work to elucidate the transcriptome profile differences between them. Specifically, we identified hub genes, performed gene ontology (GO) analysis, carried out enrichment analyses of the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, and validated hub genes using machine learning algorithms. This approach revealed a total of 136 meta-genes, 39 of which were not significant in any of the individual studies, indicating the higher statistical power of the meta-analysis. Furthermore, the results derived from the use of machine learning revealed POSTN, K35, SETD4, USP29, ANKRD37, RTN2, PRG4, and LRRC4C as substantial genes that were assigned a higher weight (0.7) than other meta-genes. Among the decision tree models, the Random Forest ones surpassed the others in adipose tissue predictive power fat deposition in fat- and thin-tailed breeds (accuracy > 0.85%). In this regard, combining meta-analyses and machine learning approaches allowed for the identification of three important genes (POSTN, K35, SETD4) related to lipid metabolism, and our findings could help animal breeding strategies optimize fat-tailed breeds' tail sizes.
Keywords: RNA-seq; fat deposition; fat-tailed sheep; machine learning.