Gene Networks from Multiple Sources of Molecular Data: Application to Livestock Species

Animal geneticists are now flooded with gigabytes of molecular data (DNA sequences; RNA expression levels; protein, metabolic, and methylation profiles) for each individual animal in a study.  Hidden within this vast mass of data are hopefully strong signals that can be exploited for livestock improvement. Gene networks offer the possibility to interpret a complex phenotype as the result of a dynamic process of interacting nodes. This contrasts with traditional approaches that compile lists with the most up- and down-regulated genes in a given experimental contrast (e.g. healthy versus disease), and DNA variants (e.g. SNP) significantly associated with the phenotype of interest based on genome-wide association studies (GWAS). Here we examine the repertoire of data resources, analytical tools and software pipelines for the inference of gene networks with particular emphasis on the application to the genetic improvement of livestock species. When the primary source is expression data, analytical methodologies based on pair-wise gene co-expression have reached some level of maturity. The adjacency function defining which genes are connected in the reconstruction of the network can be further informed by knowledge from transcription regulation and tissue-specificity. However, when the primary source is the results from GWAS, multi-variate concepts like pleiotropy and epistasis, along with a well-annotated genome become the drivers to genetically dissect complex phenotypes. To illustrate these features, we show examples of relevance to bovine, ovine and pigs production systems.