Quantitative genetic and statistical aspects of feed efficiency by analysis of the selection experiment for residual feed intake in Yorkshire pigs
The overall objectives of this dissertation were to evaluate the effects of selection for reduced residual feed intake (RFI), estimate genetic parameters, and explore different statistical and genetic models to make best use of longitudinal measurements of daily feed intake (DFI), body weight (BW), backfat (BF), and loin muscle area (LMA) to improve feed efficiency in pigs. The objectives of this dissertation were addressed in four chapters with four complementary research projects. The data used for each project were from a selection experiment for reduced RFI in Yorkshire pigs at Iowa State University. In chapter 2, a simple quadratic and linear regression of DFI and BW on age was fitted separately for individual pigs. Based on that, average daily feed intake and average daily gain were summarized for each pig. For BF and LMA, only the last measures, at off-testing, were used for each pig. By further analyses of these single summaries or measures for individual pigs, RFI had a sizable heritability of 0.29, and selection for reduced RFI was shown to have significantly decreased the amount of feed required for a given rate of growth and backfat. To find a better model to predict DFI and BW curves for individual pigs than fitting a simple quadratic and linear regression as in chapter 2, different random regression (RR) models and non-linear mixed models were evaluated on the pigs from generation 5 of the selection experiment in chapter 3. The quadratic polynomial RR model was identified to be best for DFI and BW after evaluating 40 RR models with different orders of polynomials of age and Gompertz non-linear mixed models based on predicted residual sum of squares. In chapter 4, the RR genetic analyses with quadratic order of Legendre polynomials of age and splines were applied to the data from all generations of the selection experiment to estimate genetic parameters for DFI, BW, BF, and LMA along the growth trajectory. To overcome the shortcoming of the parameters of the RR models not having a biological meaning, the hierarchical Bayesian method was applied to investigate genetic variation in the parameters of the Gompertz non-linear mixed model for DFI and BW in chapter 5. The effect of selection for reduced RFI on growth and feed intake curves was evaluated genetically by the linear RR model in chapter 4 and by the Gompertz non-linear mixed model in chapter 5. Both models showed that selection for reduced RFI has resulted in a lower feed intake curve and a lower body weight curve, especially towards the end of growth period.