Student information

Sandrine did her thesis on genomic selection in sheep

Sandrine Duchemin from Brasil finished her EMABG thesis in summer 2011 at AgroParisTech in France.

Thesis title

Genomic selection in Lacaune dairy sheep


Genomic selection yields genomic breeding values using only marker information and without looking at the phenotype of the selection candidate. Reduction of generation intervals and increasing of genetic gains in animal breeding are promising when considering genomic selection implementation. Nevertheless, outperforming traditional genetic evaluation, usually done by progeny testing of a very robust and consolidated approach, is one of the challenges of genomic selection.

In our study, we used real data from Lacaune's dairy sheep and compared the accuracies of genomic estimated breeding values (GEBV) of young males using different statistical models and methods. We studied three traits with different heritabilities. We tested three different models that considered the estimation of: a) infinitesimal effects only, b) marker effects only and c) the jointly estimation of infinitesimal and marker effects. For such different models, we also studied different methodologies,

namely BLUP, BayesC1t, Partial Least Squares (PLS) and Sparse Partial Least Squares (sPLS).

In Lacaune's dairy sheep, the inclusion of molecular information, as compared to traditional schemes, increased accuracies of estimated breeding values of young males at birth from 18 up to 25%, according to the analyzed trait. Comparisons of methods that considered molecular information did not reveal great differences in correlations among each other. In BayesC1t, the jointly estimation of marker and infinitesimal effects had slightly favourable impact on the accuracies of GEBV, specially for our low heritable trait. Inclusion of infinitesimal effects also improved slopes of predictive regression equations. In BayesC1t and sPLS, the pre-selection of markers that were retained in the final model was trait dependent. Such analysis has risen opportunities for further studies on Quantitative Trait Loci (QTL) detection, i.e. to find chromosome regions that might influence the phenotype of animals.

This work resulted in a scientific publication:

Genomic selection in the French Lacaune dairy sheep breed, S.I. Duchemin, C. Colombani, A. Legarra, G. Baloche, H. Larroque, J.-M. Astruc, F. Barillet, C. Robert-GraniƩ, E. Manfredi, Journal of Dairy Science, Volume 95, Issue 5, May 2012, Pages 2723-2733