Resumen de Network modeling for quantitative traits in pigs
Summary Pigs are important food source for humans. Moreover, they are also anatomically and physiologically similar to humans and thus serve as a good animal model for biomedical research. Traits of interest in pigs like disease resistance and meat quality are the output of complex molecular networks, which in turn are perturbed by both DNA variants and environmental factors. Understanding the molecular mechanisms underlying these traits will help improving the pig production efficiency as well as enhancing biomedical research for human health. Mapping quantitative trait loci (QTL) to elucidate the genetic basis of complex traits has been an active research field during the last decades.
Among the traits for those knowledge is scarce, despite its importance, are resistance to disease. In Chapter 2, we describe a large scale experiment aimed at identifying QTL for Enterotoxigenic Escherichia coli F41 resistance in a White Duroc × Chinese Erhualian F2 cross. The most relevant result was a highly significant QTL (genome wide P value < 0.01) around 18cM (SW2509-S0301) on pig chromosome 4 that could confer resistance to this strain.
The fast developing large scale molecular measuring technologies are providing new opportunities, but also presenting challenges in identifying molecular basis underlying the QTL effects. Among these, developing more realistic statistical models and integrating the large scale molecule data in order to explore the interplay of molecules that induce the changes of given traits, are particular interesting questions to be discussed and resolved.
In a typical model of QTL analysis, only DNA variants and a few regular covariates like sex and age are normally considered to predict a given trait. We argue that traits other the trait under study should be also added into the model. In Chapter 3, we evaluated by simulation three procedures: forward selection, undirected dependence graph (UDG), and QTL directed dependency graph (QDG) approaches to select traits other than the target trait as covariates in QTL mapping. The QDG approach appeared to have robust and better performance than other two procedures and typical QTL mapping analysis in terms of QTL detection power and false discovery rate and was applied to fatty acid composition and fat deposition traits measured in two pigs F2 crosses from China and Spain. Using the QDG approach, we identified several new QTL and potentially distinguished the indirect and direct QTL effect. This help us to reinterpret the QTL hotspot on pig chromosome 4, 6, 7 and 8 revealed previously by typical QTL mapping approach.
Finally, in Chapter 4, we studied the associations between gene expression profiles of five endocrine tissues and plasma metabolites in pigs using gene coexpression network analysis. There was limited overlap of significant transcripts and biological pathways associated with the metabolites across tissues, suggesting the associations are highly tissue specific. The significant transcripts associated with the metabolites were divided into modules, which were enriched for genes in KEGG or gene ontology categories that are related to the physiological function of the corresponding tissues. Furthermore, a multiple tissue network highlights the key role of hypothalamus genes in regulating the communication among tissues.
