International Genetic Evaluations with the BLUP method 2005
BLUP
by Þorvaldur Árnason
The BLUP method with animal model is used for genetic evaluation of Icelandic horses. The method is called BLUP (Best Linear Unbiased Prediction) since that describes the statistical properties of the method. Animal model means that all available pedigree information is used for correct weigthing of the genetic relationships among animals. The Best Linear Unbiased Prediction makes optimum use of the data on the pedigree and the scored traits and weights every piece of information correctly in accordance with the statistical model used for description of the real world. The BLUP method is able to adjust for the effects of systematic environmental faktors which can be registered in the data and make information on scores from different countries, years and on different gender comparable. The BLUP method has become a standard method for genetic evaluation of livestock (inclusive horses) worldwide.
The BLUP method with animal model for simultaneous genetic evaluation for ten correlated traits of conformation and riding ability of Icelandic horses was used as early as 1982. For several years the results were introduced to breeders by meetings and stencils with lists of results. Lists with the highest ranked horses were also published in the horse magazine Eidfaxi. The method was officially taken in use by The Farmers Association of Iceland in 1986, and since then they have been responsible for publication of results on the genetic evaluations.
In 1992 a comprehensive upgrading of the computer programs used for computing the genetic evaluations were performed and the number of scored traits incerased form ten to fourteen. Also a genetic evaluation for height at withers were added. Later genetic evaluations for mane and tail, walk and slow toelt have been added. Evaluations of the use of the BLUP method for genetic evaluations in the Icelandic population have shown that the method has been very effective and had positive impact on the Icelandic horse breeding. However, in later years an apparent need for development of a new model has become clear. The reason is that the defination of traits and form of scoring has changed over time, wich has resulted in change of genetic parameters between different periods.
International genetic evaluations of the Icelandic horse
In 2000 FEIF and the Farmers Association of Iceland assigned an agreement on establishing a data base containing a global register for the Icelandic Horse. The web application has been termed ”WorldFengur” with the domain name www.worldfengur.com. One of the main goal with WorldFengur is to establish international genetic evaluations which enables comparison of breeding horses across countries. As a first step the international genetic evaluations only include the Nordic countries but will soon be extended to include more countries where horses are scored according to FEIF standards.
During the period 1995-2001 common genetic evaluations were computed based on the Icelandic, Danish and Swedish material and the results were published in respective country for horses located in Denmark and Sweden. This could not continue due to disharmony in the data and insufficiency concerning unique identification numbers, especially in the Swedish material. Therefore in years 2002 and 2003 Danish and Swedish breeders of Icelandic horses did not have access to fresh BLUP indices. This was very unfortunate since breeders in these countries had got use to the genetic evaluations and they seemingly had positive effect on the breeding of Icelandic horses in these countries.
The construction of the database of WorldFengur lead by Jon Baldur Lorange, FAIC, offers a unique possibility for the creation of effective international genetic evaluation of the Icelandic horse. In 2003 Ágúst Sigurdsson and Thorvaldur Árnason were commissioned to develope the international genetic evaluations of the Icelandic horse to be included in WorldFengur. The project should be based on solid scientific work. The project has not been finished yet, but the fullfillment of the Nordic genetic evaluations which are now available on the web through www,worldfengur.com is an important milstone. The project has involved comprehensive statistical analyses on pedigree files and data on judging scores from Iceland and the other Nordic countries. The results were published and presented at the 55th Annual Meeting of the EAAP, Bled, Slovenia, September 5th-9th , 2004. The main results show good pedigree completedness in the data and good genetic connectedness between countries. The relationship between animals is almost the same across countries as within a country. The reason for this is the vast flow of genes from the Icelandic population. A very little difference was found between the level and variation in scores between Iceland and the other Nordic countries. This was only to be expected since Icelandic judges have been freqently represented in the judging groups in these countries. On the other hand it became clear that variation and heritability of most of the scored traits changed considerably around 1990. Therfore it seemed natural to define the same traits as two different traits depending on if they were scored before 1990 or later. Both heritability and variation increased markedly after 1989. Therfore the scores obtained before 1990 now receive less weight in the BLUP genetic evaluations. Tables with all genetic parameters used in the computation of the new BLUP indices will be published on WorldFengur. This considerable increase in the number of traits involved lead to much more complex statistical model and tricky missing patterns which required genius computing solution.
Somewhat simplified the statistical model for each trait can be written as:
Score = fixed effects of country and judging year + fixed effects of gender and age class + randomly distributed breeding values + randomly distributed envvironmental effects
All scores included in the computations are from the occation when the horse received highest notes for totalscore according to the present weighting factors. For years 1961 – 1989 the scores pertain to 14 correlated traits while for years 1990 – 2005 thirteen correlated traits are involved since willingness and character have been combined to one trait ”spirit”. Therefore in total there are
27 correlated traits simultaneusly included in the computations. In addition genetic evaluations are computed for four single traits (height at withers, mane and tail, walk and slow toelt). The BLUP indices puplished in WorldFengur are estimated breeding values for the traits as they are defined after 1990. This applies also for the horses scored earlier. Their estimated breeding values for the new traits are obtained through the genetic correlations.
In total the data for year 2004 consisted of 190,089 horses. Thereof 25,327 horses had own records divided between countries as: Iceland 20,196, Sweden 2,108, Denmark 1,678, Norway 488, Germany 466, Finland 137, The Netherlands 99, USA 57, Austria 52, UK 24 and Switzerland 23 horses.
In former genetic evaluations the BLUP indices have been scaled such that mean value for all horses with own scores were constrained to 100 and 10 index units corresponded to 1 standard deviation in the distribution. Due to the genetic progress in the population the old horses suppressed the mean of the reference group so much that the mean index for the recent part of the population was far above 100. This means that there were a kind of inflation in index values and that a devaluation was neccessary. Last year (2004) the index was completely rescaled. In the new BLUP indices the mean genetic evaluations for horses with own records obtained in the last 15 years will form the reference base. Their mean value will be constrained to 100. The distribution will be scaled such that 10 index units correspond to one genetic standard deviation for each trait. Last year the reference base group consisted of horses judged 1990-2004. The base will role each year such that this year the base consists of horses judged 1991-2005 and nect year it will be made up by horses judged 1992-2006, etc. The dispersion of the new index is slightly narrower compared with the old index.
The change in the scaling of the indices means that the level is lowered by almost five units (4 ½) compared with the old BLUP index as it was before 2004.
Information on the genetic evaluations published in WorldFengur
On www.worldfengur.com you will find following information concerning the genetic evaluations:
Height at withers (BLUP in cm as a deviation from the population mean)
Head (BLUP index)
Neck/withers/shoulders (BLUP index)
Back/loin/croup (BLUP index)
Proportions (harmony) (BLUP index)
Legs (BLUP index)
Leg stance (correctness) (BLUP index)
Hoofs (BLUP index)
Mane and tail (BLUP index)
Toelt (BLUP index)
Slow toelt (BLUP index)
Trot (BLUP index)
Pace (BLUP index)
Gallop (BLUP index)
Spirit (BLUP index)
Form under rider (expression) (BLUP index)
Walk (BLUP index)
Index for body conformation
Index for riding ability traits
Index for total score
Country of scoring (IS, DK, etc.)
Number of parents registered (2, 1 or 0)
Total number of progenies registered when the genetic evaluations were computed
Number of progenies with complete scores
Number of progenies with height at withers measured
Number of progenies with a score for mane and tail
Number of progenies with a score for slow toelt
Number of progenies with a score for walk
Accuracy of the genetic evaluation (Correlation between the true and the estimated breeding value in %)
Standard error of the estimate
Inbreeding coefficient in %
Information about the impact of offspring on the index for conformation, riding ability and totalscore, as a deviation from an index based on ancestors and the individuls's own scores, will be published on WorldFengur later. A new computer program for such calculations is currently under development.
November 2005,
Thorvaldur Árnason