| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
,1
* INRA, UR337 Station de Génétique Quantitative et Appliquée, F-78350 Jouy en Josas, France
Institut de l’élevage, 149 rue de Bercy, 75595 Paris Cedex 12, France
Union nationale des coopératives d’élevage et d’insémination animale, 149 rue de Bercy, 75595 Paris Cedex 12, France
1 Corresponding author: Francois.Guillaume{at}jouy.inra.fr
 |
ABSTRACT |
|---|
 TOP ABSTRACT ACKNOWLEDGEMENTSREFERENCES |
|---|
Key Words: dairy cattle • marker-assisted selection
Over the last decade, several QTL detection programs have been conducted in dairy cattle (e.g., Georges et al., 1995; Boichard et al., 2003). Such studies revealed the existence of several QTL with large effect on dairy traits such as the gene encoding acylCoA:diacyglycerol acyltransferase (DGAT1) on chromosome 14 (Grisart et al., 2002), the ATP-binding cassette, subfamily G, member 2 gene (ABCG2) on chromosome 6 (Cohen-Zinder et al., 2005), or the growth hormone receptor (GHR) on chromosome 20 (Blott et al., 2003). Use of these QTL in a marker-assisted selection (MAS) program has the potential to improve selection efficiency in dairy cattle (Kashi et al., 1990). Since 2001, such a MAS program has been implemented in France in the Holstein, Normande, and Montbéliarde breeds (Boichard et al., 2002). One of the objectives of this program is to help breeding companies select which young bulls should be progeny tested. The accurate genetic value of selected animals can only be calculated after progeny testing (approximately 5 yr after the selection decision has been made). Consequently, efficiency of the MAS program is difficult to prove on real data and can be estimated only after a few years of implementation. In January 2007, the accuracy of production traits’ EBV of young bulls born in 2001 was high because of the records of their progeny daughters. These bulls were also genotyped for MAS and can therefore be used to assess the precision of MAS breeding values. The objective of this study was to check, using real data, if breeding values estimated for young animals using MAS were more precise than breeding values obtained at the same age based only on a polygenic model.
Files of the evaluation of April 2004 (the oldest conserved MAS data set) were used in this study. This evaluation used a pedigree of 34,318 animals of which 23,137 had phenotypic records and 16,629 were genotyped. Animals were genotyped for 43 microsatellite markers, which were used to follow the transmission of 14 QTL regions that relied on 2 to 5 evenly spaced microsatellite markers. These QTL were selected on basis of the results of a QTL detection program (Boichard et al., 2003) and confirmed with new sire families (Druet et al., 2006).
Phenotypic records were twice the daughter yield deviations (DYD) for males and yield deviations for females computed for milk, fat, and protein yields and fat and protein percentages, pooled from the first 3 lactations jointly as in VanRaden and Wiggans (1991). These records were obtained from the official genetic evaluations of both April 2004 and January 2007 (Robert-Granié et al., 1999). Respective weights were estimated as in VanRaden and Wiggans (1991) with a correction for number of cows in each herd.
The model used in this study was a single-trait and multiple-QTL model as proposed by Fernando and Grossman (1989):
 | [1] |
where y is a vector containing records, β is a vector of fixed effects (the mean), u is a vector of random polygenic effects, vi is a vector of random gametic effects for QTL i, n_qtl is the number of QTL considered for the trait, and e is a vector of random residual terms. X, Z, and Zvi are known design matrices relating results to fixed, random polygenic, and gametic effects, respectively. Probability of identity-by-descent matrices (Zvi) were obtained using a method similar to that of Pong-Wong et al. (2001). Between 4 and 5 QTL were used for each production trait, and the variance components (see Table 1
) were assessed in a previous study (Druet et al., 2006).
|
) were compared. These correlations were obtained for milk, fat, and protein yields and fat and protein percentages.
|
A simulation study performed on the French MAS program (Guillaume et al., 2008) estimated the expected gains of correlation from MAS EBV over pedigree index and also their variation. The present results are in agreement with the results obtained by simulation: correlations are slightly lower for milk yield and protein yield and content, and greater for fat yield and percentage. The simulation study showed that these gains are expected to vary for each crop of bulls and that use of DYD instead of the unknown true genetic values underestimates the gain in accuracy achieved by MAS. Finally, Guillaume et al. (2008) indicated that the gain in accuracy achieved by MAS is clearly larger for young bulls born in 2006 than for animals born during the first years of the program. Indeed, the program is accumulating more information; approximately 10,000 new genotyped animals are integrated into the database each year. These additional genotypes improve the efficiency of the program.
To further increase the efficiency of the program, breeding companies have decided to genotype dams of young bulls and some progeny daughters of sires of young bulls. This targeted genotyping has been shown to enhance the efficiency of MAS. In the batch of 899 candidates studied, only 66% of the candidates’ dams were genotyped and first-crop daughter genotypes of only 6 of 39 sires were available, whereas these numbers are now much larger. Finally, the set of microsatellite markers was also changed in 2005, and the QTL are now followed with more accuracy.
In this study, the greatest increases in terms of correlation were observed for milk content traits. This can be explained by the fact that a larger proportion of the genetic variance was explained by the QTL for these traits. The lowest gain was obtained for protein yield, in which QTL explained only 35% of the genetic variation. Gains in correlation are ordered in the same way as proportion of genetic variance explained by QTL, so that the proportion of genetic variance explained by QTL should be large enough to obtain improvement of accuracy. Furthermore, content traits were influenced by QTL for which average informativity weighted by proportion of variance explained by each QTL was greater (Table 1
). This is partly due to the low informativity achieved with the markers for QTL on Bos taurus autosomes 19 and 26, which influenced yield traits. To resolve this problem, new microsatellite markers were selected in 2005.
In the present data set, EBV were available only for progeny-tested bulls. Candidates not selected for progeny testing were not included in the study. These animals correspond to those candidates that had poor MAS EBV. A study including all candidates would better assess the efficiency of MAS. Unfortunately, precise genetic values based on progeny daughters are not available for unselected candidates.
In the coming years, efficiency of MAS is expected to improve because our knowledge of the genome is increasing. Methods using genetic markers in animal selection will certainly change. Indeed, future methods will use markers closer to the QTL or use the mutations directly responsible for QTL variation. Linkage disequilibrium between markers and QTL can be used in these conditions. Finally, because of the ability to genotype animals at many more markers at a reasonable cost, genomic selection (Visscher and Haley, 1998; Meuwissen et al., 2001) will be possible.
|
ACKNOWLEDGEMENTS |
|---|
|
TOPABSTRACTACKNOWLEDGEMENTSREFERENCES |
|---|
Received for publication November 5, 2007. Accepted for publication February 8, 2008.
|
REFERENCES |
|---|
|
TOPABSTRACTACKNOWLEDGEMENTSREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
