Genomic selection for residual feed intake

© 2016 Dr. Majid KhansefidThe efficiency of animals in converting feed to products is one of the main factors influencing the profitability of beef and dairy production. However, inclusion of feed efficiency as a new trait into the breeding objective requires re-estimation of all economic weights for other traits unless it is independent of them. A convenient measure of efficiency is residual feed intake (RFI) which is the difference between actual intake and predicted feed intake and hence is independent of other traits used in the prediction. RFI and other measures of feed efficiency are expensive to measure because they requires precise measurements of individual feed consumption and this has limited direct selection for feed efficiency in cattle. Genomic selection, using single nucleotide polymorphisms (SNP) genotypes to estimate breeding values without measuring feed intake on selection candidates, could overcome this limitation. However, genomic selection requires a reference population that has been measured for RFI and genotyped. The limited size of reference populations, due to the high costs for RFI measures, currently results in modest prediction accuracies. The aim of the research in this thesis is to increase the accuracies of genomic estimated breeding values (GEBVs) for RFI and to find biological processes and genes associated with RFI variation using genomic data (RNA and DNA). Chapter 1 reviews methods of selecting for RFI and concludes that genomic selection offers the best opportunity. Chapter 2 combines data from multiple breeds to increase the size of the reference population. We show that there is substantial SNP × breed interaction variance but, despite this, the combined reference increases the accuracy of GEBV especially for breeds with a small number of animals in the reference. An alternative analysis used the beef cattle data to identify SNPs with a larger than average effect on RFI and gave these SNPs extra weight in calculating GEBVs for Holstein cattle. This increased the accuracy of GEBV in Holsteins. The results showed the highly polygenic nature of RFI. However, to get most benefit from a multiple breed reference population, it is necessary to use methods of analysis that concentrate on SNPs near causal variants. Therefore, an effort was made to find genes and SNPs that have an effect on RFI using gene expression measured by RNA sequencing data (RNA-Seq) from Angus bulls (muscle and liver samples) and Holstein cows (liver and blood samples). Chapter 3 reports many genes whose expression was correlated with RFI. These genes were represented in wide range of biological processes such as energy metabolism and proteolysis. However, correlations do not prove causation and there is no guarantee that SNPs in these genes influence RFI. SNPs might influence RFI by changing the expression of the genes, so in Chapter 4, RNA-Seq data was analysed to find cis acting expression quantitative trait loci (eQTL). A traditional local eQTL analysis with allele specific expression (ASE) was compared and both methods identified the same cis eQTL in many cases although there were systematic differences, due for instance to parent of origin effects. There was a significant overlap between SNPs associated with gene expression and SNPs associated with traditional phenotypes including RFI. This shows that some QTL are actually eQTL. In chapter 5, by combining the results from other chapters, genes were found where a SNP was associated with both RFI and expression of a gene whose expression was correlated with RFI. These genes and SNPs are good candidates for use in genomic selection but the effects are mostly small and need to be verified in future research to discover their actual biological function in RFI variation before they are used with high confidence in genomic selection