OATAO - Open Archive Toulouse Archive Ouverte Open Access Week

New genetic longitudinal models for feed efficiency

Tran, Van hung. New genetic longitudinal models for feed efficiency. PhD, Pathologie, Toxicologie, Génétique et Nutrition, Institut National Polytechnique de Toulouse, 2018

(Document in English)

PDF (Author's version) - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader


Although non-genetic and genetic approaches heavily improved feed efficiency in the last decades, feed cost still contributes to a large proportion of pork production costs. In addition, thelimited effi-ciency of feed use not only increases the environmental impact due to the waste of feed. Over the last decades, advances in high-throughput technologies for animal management,including automat-ic self-feeders, created a proliferation of repeated data or longitudinal data. The objective of this thesis was to develop new genetic models to better quantify the genetic potentialof animals for feed efficiency using longitudinal data on body weight (BW), feed intake and body composition of the animals. Data from 2435 growing Large White pigs from a divergent selectionexperiment for resid-ual feed intake (RFI) were used. In this population, males were weighted every week whereas fe-males and castrated males were weighted every month at the beginning of the test (10 weeks of age) and more often towards the end of the test (23 weeks of age). In a first step, different approaches investigated how to predict missing weekly BW for intermediate stages. For the tested period, a quasi linear interpolation based on the adjacent weeks is the best approach to deal with missing BW in our dataset. In a second step, different longitudinal models, such as random regression (RR) mod-els, structured antedependence models (SAD) and character process models, in which the covari-ances between weeks are accounted for, were compared. The comparison focused on best-fit to the data criteria (Loglikelihood, Bayesian Information Criterion), on variance components estimations (heritability estimates, genetic variances and genetic correlations between weeks) and on predictive ability (Vonesh concordance coefficients). The results showed that SAD is the most parsimonious model for feed conversion ratio (FCR) and for RFI, two measures of feed efficiency. The SAD model also provided similarpredictive abilities as the other models. A selection criterion combining the weekly breeding values was proposed for practical applications to selection. In addition, we evaluated the potential of genomic information to improve the accuracy of breeding value predictions for aver-age daily gain and residual feed intake, applying single step genomic approaches to the RR and SAD models. In our dataset, prediction accuracies was low for both traits, and was not much improved by genomic information. Finally, we showed that divergent selection for RFI had a major impact on the FCR and RFI profile trajectories in each line. In conclusion, this thesis showed that selection for trajectories of feed efficiency is feasible with the current available information. Further work is needed to better evaluate the potential of genomic information with these models, and to validate strategies to select for these trajectories in practice

Item Type:PhD Thesis
Uncontrolled Keywords:
Institution:Université de Toulouse > Institut National Polytechnique de Toulouse - Toulouse INP (FRANCE)
Laboratory name:
Research Director:
David, Ingrid and Gilbert, Hélène
Deposited On:14 Apr 2020 11:15

Repository Staff Only: item control page