Advantages of BLUP animal model for breeding value estimation in horses
Abstract
In order to benefit from the theoretical advantage of the BLUP animal model, an across-herd evaluation system was developed which reduces computation cost for a routine application to a large horse population (80 000 horses and 4000 stallions).
The model used for the log of annual earnings includes an individual additive genetic component, permanent environmental and herd × dam effects considered as random and sex age × year subclass effects as fixed. This herd × dam effect is caused by the difference between the maternal and paternal components and represents the common environmental conditions shared by progeny of the same mother. The variance-covariance matrix includes the complete relationship matrix between all the animals whatever the kind of animal: recorded or non-recorded; parent or non-parent horses; males or females.
The system of mixed-model equations is solved by using a particular structure of the inverse of the relationship matrix. Animals with no progeny and mares are absorbed. The resulting system (fixed effects and stallions) is solved by iteration. An approximation of the coefficient of determination is then calculated by using matrix approximations.
This method is of very great practical interest in horse populations, where rigorous selection plans are rare.
Résumé
Pour bénéficier des avantages théorique du BLUP en modèle individuel nous avons mis au point une indexation en routine d'une population importante (80 000 chexaux pour 4000 étalons) avec des coûts informatiques réduits.
Le modèle utilisé pour le logarithme du gain annuel comprend une composante valeur génétique additive individuelle, un effet de milieu commun aux différentes performances et un effet “maternel-élevage” considér'es comme aléatoires ainsi que des effets sexe et classe d'age-année de performances considérés comme fixes. L'effet “maternel-élevage” correspond à l'écart observé entre la composante maternelle et paternelle de la variance et se définit par les effets de milieu communs aux descendants d'une même mère.
La matrice de variance-covariance comprend la matrice de parenté, quelle que soit la nature des individus: performeur ou non, reproducteur ou non, male ou femelle.
Le systéme d'équations du modèle mixte est résolu grace à la structure particulière de l'inverse de la matrice de parenté. Les animaux non-reproducteurs et les poulinières sont absorbées. Le syst`eme final effets fixes et mals reproducteurs est resolu par itération. les coefficients de détermination sont calculés de façon approchée par le calcul matriciel.
La méthode apporte beaucoup d'avantages pratiques pour les populations chevalines qui manquent le plus souvent de plans de sélection rigoureux.
Zusammenfassung
Um die theoretischen Vorteile des BLUP-Animal-Model praktisch ausnutzen zu können, wurde ein herdenübergreifendes Auswertungssystem entwickelt, das den Rechenaufwand füie Routineanwendung in einer groβen Pferdepopulation (80 000 Pferde und 4000 Hengste) reduziert.
Das Modell zur Auswertung der logarithmierten jährlichen Gewinnsumme berücksichtigt eine individuelle additiv-genetische Komponente, permanente umweltbedingte und “Herde × Mutter”-Effekte als zufällig und Geschlecht und Alter × Jah Effekte als fix. Der “Herden × Mutter”-Effekt tritt wegen der Differenzen zwischen maternalen und paternalen Komponenten auf und stellt die gemeinsamen Umweltbedingungen von Nachkommen der gleichen Mutter dar. Die Varianz-Kovarianz-Matrix umfaβt die gesamte Verwandtschaftsmatrix zwischen allen Tieren, unabhängig ob die Tiere mit oder ohne Daten sind, als Elterntiere oder nicht auftreten, oder männlich oder weiblich sind.
Das System der Mixed-Model-Gleichungen wird gelöst, indem eine besondere Struktur der invertierten Verwandtschaftsmatrix genutzt wird. Tiere ohne Nachkommen und Stuten werden absorbiert. Das entstehende System (fixe Effekte und Hengste) wird iterativ gelöst. Das Bestimmtheitsmaβ wird approximativ berechtnet.
Diese Methode ist von seh groβer praktischer Bedeutung in Pferdepopulationen, in denen strikte Selektionspläne selten angewendet werden.
References (18)
- C.R. Henderson
Use of relationships among sires to increase accuracy of sire evaluation
J. Dairy Sci.
(1975) - C.R. Henderson
Use of all relatives in intra-herd prediction of breeding values and real producing abilities
J. Dairy Sci.
(1975) - L.D. Van Vleck et al.
Successive overrelaxation, block iteration and method of conjugate gradients for solving equations for multiple trait evaluation of sires
J. Dairy Sci.
(1985) - C.R. Henderson
Best linear unbiased estimation and prediction under a selection model
Biometrics
(1975) - C.R. Henderson
A simple method for computing the inverse of a numerator matrix used in prediction of breeding values
Biometrics
(1976) - G.F.S. Hudson et al.
Genetic evaluation of swine for growth rate and backfat thickness
J. Anim. Sci.
(1985) - B. Langlois
Analyse statistique et génétique des gains des Pur Sang Anglais de trois ans dans les courses plates françaises
Ann. Génét. Sél. Anim.
(1975) - B. Langlois
Estimation de la valeur génétique des chevaux de sport d'après les sommes gagnées dans les compétitions équestres françaises
Ann. Génét. Sél. Anim.
(1980) - B. Langlois
Evaluation génétique des Trotteurs Français d'après leurs gains
Journée de la recherche chevaline organisée par le CEREOPA
(1986)
Cited by (17)
Genetic approaches for increasing fitness in endangered species
2022, Trends in Ecology and EvolutionThe global rate of wildlife extinctions is accelerating, and the persistence of many species requires conservation breeding programs. A central paradigm of these programs is to preserve the genetic diversity of the founder populations. However, this may preserve original characteristics that make them vulnerable to extinction. We introduce targeted genetic intervention (TGI) as an alternative approach that promotes traits that enable species to persist in the face of threats by changing the incidence of alleles that impact on fitness. The TGI toolkit includes methods with established efficacy in model organisms and agriculture but are largely untried for conservation, such as synthetic biology and artificial selection. We explore TGI approaches as a species-restoration tool for intractable threats including infectious disease and climate change.
Inheritance of racing performance of trotter horses: An overview
2009, Livestock ScienceHarness racing is a form of horseracing in which the horses race in a specified gait (trot or pace). In contrast to the Thoroughbred, the trotter is not an international breed. In this type, the horses are raced with trotting or pacing gait. Breeds specialized for racing at trot or pace are indigenous to many countries. Separate breeds of light harness horses, generally designated as trotters have evolved for racing purposes in several countries. The important horse breeds used for harness racing in different parts of the world are the Standardbred, French Trotter, Swedish Trotter, Orlov Trotter, Russian Trotter, Finnhorse, Icelandic Toelter, Dole horse and North-Swedish cold-blooded horses. The trotter ranks worldwide second to the Thoroughbred in popularity as a racehorse. Racing performance in trotters, in contrast to Thoroughbred is characterized by qualifying tests before entering the races, inclusion of more than one breed in international races only and greater duration of racing career. An intensive selection of stallions on the basis of phenotypic racing performance has been practised in many trotter populations for quite a long time. Unlike Thoroughbreds, improvements have been observed in different trotter populations and this is attributed to both genetic and environmental changes. Environmental changes include enhanced training methods, as well as improved tracks, harness and sulkies. As a result of selection, racing time of trotters has been reduced over the years. The estimated annual genetic progress in racing performance traits of Swedish Standardbred horses corresponds to 5% of the phenotypic standard deviation, 3.6% in French Trotters and 5% in Dutch Trotters. According to the recent selection for speed in trot, this trait remains heritable and genetic improvement is observed in most countries. Correlations between earnings and times are negative and high, and hence favourable. As a result, selection based on times and earnings are quite effective. A multiple trait approach avoids potential biases of one particular measure, even if the objective of all traits is much the same. Since racing performance may be evaluated in both males and females and repeated observations can be obtained on the same animal, mass selection based on performance tests would be the selection procedure of choice. In the future, interest in the possible use of marker-assisted selection (MAS) for enhanced genetic improvement in horses is likely to increase. MAS is likely to be a valuable complement to selection of horses based on estimated breeding values (EBVs) obtained by the Best Linear Unbiased Prediction (BLUP) method, rather than as a replacement for EBVs.
Genetic study of gestation length in andalusian and arabian mares
2006, Animal Reproduction ScienceThe length of gestation in Andalusian, or Spanish Purebred (SPB) and Arabian (AB) mares reared in Spain was analysed, based on 766 spontaneous full-term deliveries appertaining to 141 mares of SPB breed and 72 mares of AB breed in 31 breeding seasons. The data were obtained from the Yeguada Militar de Jerez de la Frontera stud farm in Cádiz, Spain. The mean length of gestation was of 336.8 ± 0.48 days in the SPB mares and 340.3 ± 0.63 days in AB mares. To assess the accurate prediction of time of birth the potential effect of a number of factors was investigated. The influences of the breed, mare, month and year of mating, age of the mother, number of births and sex of the foal were statistically significant. The factor have the greatest influence over the gestation length was the mare itself, with a correlation among consecutive births of around 0.4. The effect of inbreeding, both of the mare and foal, was negligible. Gestation length shortened as the breeding season progressed: in both breeds, a delay of 1 month in mating corresponded to a decrease of 3 days in the gestation length. According to our results, gestation length decrease as the mare gets older, with the shortest gestation periods when the mare is 10–12 years old, and from this point on, it slowly increases. The gestation period shortens as the 4th or 5th birth approaches, and then gets progressively longer. The range of variation in gestation length due to the number of births to the mare is of 2.9 days for the AB mares, and 2.2 days for SPB mares. The heritability for the gestation length for AB and the SPB breeds was 0.2, with a repeatability of 0.36 and 0.37, for SPB and AB breeds, respectively. With the data from both breeds, and using a classical approach, the response to selection was estimated if mares with extreme gestation lengths were culled, i.e. lengths which are under 310 days, or over 360 days. According to our results, in the case of SPB, a decrease of 14–45% would occur in the number of extreme gestation lengths, while in the AB breed, this value would decrease from 2 to 39%.
Practical efficiency of breeding value estimations based on annual earnings of horses for jumping, trotting, and galloping races in France
2004, Livestock Production ScienceIn France, breeding value estimations for horses are calculated according to a BLUP animal model fitted to the log of annual earnings in jumping, trotting races, flat races, steeplechases and hurdle races.
The regression bp/I0 and correlation R between the breeding value estimation at the moment of conception with the mean of future performances P make it possible to check the practical efficiency of the method of indexation used.
The expected regression is one and the expected correlation depends on heritability and the mean of the determination coefficient of the index I0. The results are the following.
The situation appears to be suboptimal in the case of jumping but improvements are certainly possible in the case of races. The obtained results, however, confirm the usefulness of these tools for trotting and flat races. The case of steeplechases and hurdle races is not as effective mainly because of the low precision of the estimations. Other predictors of performance, BLUP for yearlings or at 3 years of age for jumpers, confirm this first analysis and the results, age per age, are detailed in the text.
Efficacité pratique des indices de sélection fondés sur les gains annuels en concours hippique et dans les courses de trot et de galop en France.
L'estimation de la valeur génétique des chevaux est calculée en France par un BLUP en modèle animal ajusté sur le logarithme des gains annuels en concours hippique, en courses au trot et au galop, en plat et à l'obstacle.
La régression bp/I0 et la corrélation R entre l'estimation de la valeur génétique au moment de la conception, avec la moyenne des futures performances P autorise à vérifier l'efficacité théorique et pratique de la méthode d'indexation mise en œuvre. La régression attendue est de 1 et la corrélation attendue dépend de l'héritabilité et du coefficient de détermination moyen de l'indice I0. Les résultats obtenus sont les suivants:
La situation apparaı̂t être suboptimale dans le cas du concours hippique mais des améliorations sont très certainement envisageables dans le cas des courses. Toutefois, les résultats obtenus confirment l'utilité de ces outils de sélection pour les courses au trot et les courses plates. Dans le cas des courses à obstacles c'est moins évident principalement à cause de la faible précision moyenne des estimations. D'autres prédicteurs de la performance, le BLUP des yearling ou des 3 ans pour les chevaux de sport confirment cette première analyse et les résultats âge par âge sont détaillés dans le texte.
Mots clés: Cheval/Valeur génétique/Gains/BLUP/Modèle animal/Prédiction des performances/Compétitions équestres/Courses
Application of standardized, accumulated transformed earnings in breeding of Norwegian trotters
1994, Livestock Production ScienceEarnings of horses were accumulated over age-class(es), and unraced horses were assigned zero earnings. The variables were power transformed (0.20) and standard normalized within birth year. Four resulting accumulated transformed and standardized earnings (ATSE) variables, for 1890 offspring by 74 base population sires, were analysed. A multivariate sire model was used in an initial analysis aimed at estimating phenotypic parameters, while a corresponding mate-corrected sire model was applied to estimate genetic parameters. Breeding values of mates were predicted by a multivariate animal model using data of all animals (14054) and parameter estimates from the initial analysis. The heritability estimates of ATSE variables over age-classes 3, 3–4, 3–5 and 3–6 year increased with increasing percent of raced horses and were 0.14, 0.20, 0.23 and 0.22, respectively. Phenotypic correlations were all greater than 0.70, and the genetic correlations approached 1.00. Sex and birth year effects were included in all analyses and in the proposed 4-trait animal model evaluation routine. Young sires should be selected on ATSE3-4 breeding values which showed a significant genetic trend of 3.2% of a phenotypic standard deviation per year.
Les gains des chevaux ont été cumulés en fonction des classes d'âge, et les chevaux sans course se sont vus attribuer un gain nul. Les variables ont étéélevées à la puissance 0.20 puis standardisées intra année de naissance. Quatre variables de gains cumulés, transformés et standardisés (ATSE) ont été analysés. Un modèle père multivariate a été utilisé dans une analyse initiale visant à estimer les paramètres phénotypiques, tandis qu'un modèle père corrigé pour les conjointes était utilise pour l'estimation des paramètres génétiques. Les valeurs génétiques des conjointes ont été prédites selon un modèle animal multivariate utilisant les données de tous les animaux (14054) et les estimées de parametres de la première analyse. Les estimées d'héritabilité des variables ATSE pour les âges de 3, 3–4, 3–5, 3–6 ans ont augmenté avec le pourcentage croissant de chevaux participant aux courses et ont été de 0.14, 0.20, et 0.23 et 0.22 respectivement. Les corrélations phénotypiques ont toutes été supérieures à 0.70 et les corrélations génétiques ont été proches de 1. Les effets du sexe et de l'année de naissance ont été inclus dans toutes les analyses et dans le modèle animal à 4 caractères proposés en routine. Les jeunes pères doivent être sélectionnés sur estimées de valeur génétique pour ATSE 3-4, ce qui conduirait à un progrès génétique annuel de 3,2% d'écart type phénotypique.
A Pseudo-Absorption Strategy for Solving Animal Model Equations for Large Data Files
1990, Journal of Dairy ScienceIterative Gauss-Seidel or Jacobi methods have been extensively used worldwide to solve mixed model equations arising in genetic evaluation, but convergence rates have been very low and central memory requirements high for most applications of animal models with large data files. A new computing algorithm is proposed that is especially adapted to dairy cattle evaluation using all lactations, with group effects for unknown parents only, and without stringent limitations on the number of fixed effects included in the model. The iterative procedure takes advantage of the known structure of the coefficient matrix after absorption of permanent environmental effects. The proposed strategy, although more complex to implement than classical iterative methods, does not require computers with extremely large central memory and appears to have a satisfactory convergence rate. The method is illustrated by a numerical example dealing with milk yield evaluation of the Normande breed (2,019,137 females).