Integration/blending of external information into genetic/genomic evaluations

Diverse speeche and writing (2012)

Mid-infrared prediction of milk titratable acidity and its genetic variability in first-parity cows

Conference (2012, August 27)

Coagulation of milkhas a direct effect on cheese yield. Among several parameters, titratable acidity of milk (TA) influences all the phases of milk coagulation. In order to study the genetic variability ... [more ▼]

Coagulation of milkhas a direct effect on cheese yield. Among several parameters, titratable acidity of milk (TA) influences all the phases of milk coagulation. In order to study the genetic variability of this trait on a large scale, mid-infrared (MIR) chemometric methods were used to predict TA. A total of 507 milk samples collected in the Walloon Region of Belgium from individual cows were analyzed using a MIR spectrometer. TA was recorded as Dornic degree. An equation to predict TA from milk MIR spectrum was developed using partial least squared regression after a first derivative pre-treatment applied to the spectra to correct the baseline drift. During the calibration process, 45 outliers were detected and removed from the calibration set. The TA mean of the final calibration set was 16.62 (standard deviation (SD) = 1.80). The coefficient of determination (R²) was 0.82 for the calibration with a standard error (SE) of 0.76. A cross-validation (cv) was performed (R²cv = 0.81 with SEcv = 0.80). This equation was then applied on the spectral database generated during the Walloon routine milk recording. The variances components were estimated by REML using single-trait random regression animal test-day model. The dataset used included 33,717 records from 9,191 Holstein first-parity cows; the TA mean was 17.05 (SD = 1.35) and TA ranged from 12.83 to 20.87. Estimated daily heritabilities ranged from 0.43 at 5th day in milk to 0.59 at 215th day in milk indicating potential of selection. Further research will study phenotypic and genetic correlations between TA and milk production traits. [less ▲]

Mid-infrared prediction of milk titratable acidity and its genetic variability in first-parity cows

in Book of Abstracts of the 63rd Annual Meeting of the European Federation of Animal Science (2012, August)

Coagulation of milkhas a direct effect on cheese yield. Among several parameters, titratable acidity of milk (TA) influences all the phases of milk coagulation. In order to study the genetic variability ... [more ▼]

Coagulation of milkhas a direct effect on cheese yield. Among several parameters, titratable acidity of milk (TA) influences all the phases of milk coagulation. In order to study the genetic variability of this trait on a large scale, mid-infrared (MIR) chemometric methods were used to predict TA. A total of 507 milk samples collected in the Walloon Region of Belgium from individual cows were analyzed using a MIR spectrometer. TA was recorded as Dornic degree. An equation to predict TA from milk MIR spectrum was developed using partial least squared regression after a first derivative pre-treatment applied to the spectra to correct the baseline drift. During the calibration process, 45 outliers were detected and removed from the calibration set. The TA mean of the final calibration set was 16.62 (standard deviation (SD) = 1.80). The coefficient of determination (R²) was 0.82 for the calibration with a standard error (SE) of 0.76. A cross-validation (cv) was performed (R²cv = 0.81 with SEcv = 0.80). This equation was then applied on the spectral database generated during the Walloon routine milk recording. The variances components were estimated by REML using single-trait random regression animal test-day model. The dataset used included 33,717 records from 9,191 Holstein first-parity cows; the TA mean was 17.05 (SD = 1.35) and TA ranged from 12.83 to 20.87. Estimated daily heritabilities ranged from 0.43 at 5th day in milk to 0.59 at 215th day in milk indicating potential of selection. Further research will study phenotypic and genetic correlations between TA and milk production traits. [less ▲]

Extension of Bayesian procedures to integrate and to blend multiple external information into genetic evaluations

Conference (2012, July 18)

Current genetic evaluations are mostly based on local data, potentially followed by an international second step, as it is performed by INTERBULL (Uppsala, Sweden) for dairy breeds. However, reliability ... [more ▼]

Current genetic evaluations are mostly based on local data, potentially followed by an international second step, as it is performed by INTERBULL (Uppsala, Sweden) for dairy breeds. However, reliability of estimated breeding values (EBV) for animals with few local data may be limited. Furthermore, the current development of genomic selection makes this issue more important and the blending of those multiple sources more necessary. Current methods are based on selection index and Bayesian procedures were proposed for a single external source. The aim of this research was to extend these methods to integrate information from multiple sources outside a given evaluation system, i.e. external information, into this genetic evaluation, i.e. internal evaluation, and to assess its potential to blend multiple sources of information when only external information is available. To allow simplifications of the computational burden and the propagation of all external information through the whole pedigree, all animals were considered as animals associated to external information thanks to the prediction of external information from available one for internal animals, i.e. animals with no external information. Multiple considerations of contributions due to relationships among animals were also taken into account. Two dairy cattle populations were simulated across 5 generations. Internal females were randomly mated with internal and 50 external males. Milk production for the first lactation for each female was simulated in both populations. Results for 100 replicates showed that average rank correlations among Bayesian EBV and EBV based on the joint use of external and internal data were close to 1 for both external and internal animals. The respective correlations for the internal evaluation were equal to 0.54 and 0.95. Mean squared error, expressed as a percentage of the internal mean squared error, was close to 0% for both external and internal animals. Thereby, the Bayesian procedure has the potential to integrate and to blend multiple sources of external information to evaluate animals more reliably. [less ▲]

Rapport d'activités final du premier mandat (du 1er janvier 2010 au 31 décembre 2011) : DairySNP, Etude de la variabilité génomique des bovins laitiers et mixtes en vue de mieux connaître leur biodiversité et d'initier une sélection génomique au sein de ces races

Report (2012)

Integration of external information into genetic evaluations by a Bayesian procedure

Poster (2012, February 10)

One of the most important theoretical assumptions of methods used to assess genetic values is that all available information has to be considered simultaneously to obtain unbiased estimates. However, the ... [more ▼]

One of the most important theoretical assumptions of methods used to assess genetic values is that all available information has to be considered simultaneously to obtain unbiased estimates. However, the widespread international exchange of genetic material and, more recently, the important development of the genomic selection lead to the coexistence of different genetic evaluations. Therefore, the blending of the different sources of information is necessary to achieve better prediction. Integration of external information into genetic evaluations by a Bayesian procedure can partially resolve the problem under some assumptions. Results from such a method that also avoids double counting among external animals are highly similar to those from a joint evaluation. [less ▲]

Mid-Infrared Prediction of Lactoferrin Content in Bovine Milk: Potential Indicator of Mastitis

in Animal (2012), 6

Extension of Bayesian procedures to integrate and to blend multiple external information into genetic evaluations

in Journal of Dairy Science (2012), 95(Supplement 2),

Integration of foreign breeding values for stallions into the Belgian genetic evaluation for jumping horses

in Book of Abstracts of the 62nd Annual Meeting of the European Federation of Animal Science (2011, August)

The aim of this study was to test the integration of foreign estimated breeding values (EBV) for stallions into the Belgian genetic evaluation for jumping horses. Belgian breeders import horses from ... [more ▼]

The aim of this study was to test the integration of foreign estimated breeding values (EBV) for stallions into the Belgian genetic evaluation for jumping horses. Belgian breeders import horses from neighbouring countries for which foreign information is needed as prior to estimate a more accurate EBV. The Belgian model is a bivariate repeatability BLUP animal model. For the year 2009, pedigree and data files contained 101 382 horses and 712 212 performances. 98 French and 67 Dutch stallions were selected and their foreign EBVs were converted into Belgian national trait. Associated reliabilities were also estimated. A Bayesian approach (Legarra et al., 2007) was applied to integrate this prior information into the Belgian evaluation. It led to a slight modification of the average EBV and of the standard deviation for the whole population. It also led to a new Belgian ranking of the foreign stallions more similar to foreign rankings. However, the adequacy of the Belgian model was not damaged. With regards to prediction ability, Bayesian evaluations using conversion equation estimated by Weighted Least Squares procedure predicted the best traditional EBVs of the year 2009 for the French stallions. For the Dutch ones, it were the evaluations associated to the conversion equation based on Wilmink et al. (1986). For both countries, Bayesian evaluations using conversion equation based on Goddard (1985) improved the most the stability of EBVs. Finally, integration of French and Dutch information improved reliabilities of the Bayesian EBVs of at least 5% and 2% for French and Dutch stallions, respectively. These results confirm the interest to integrate foreign information into the Belgian evaluation for jumping horses. [less ▲]

Integration of foreign breeding values for stallions into the Belgian genetic evaluation for jumping horses

Master's dissertation (2010)

The aim of this study was to test the integration of foreign breeding values for stallions into the Belgian genetic evaluation of jumping horses. Belgian breeders import horses from neighbouring countries ... [more ▼]

The aim of this study was to test the integration of foreign breeding values for stallions into the Belgian genetic evaluation of jumping horses. Belgian breeders import horses from neighbouring countries for which foreign information is needed as prior to estimate a more accurate breeding value. The Belgian model is a bivariate repeatability BLUP animal model. For the year 2009, pedigree and data files provided by the Catholic University of Leuven contained 101 382 horses and 712 212 performances. 98 French and 67 Dutch stallions were selected and their foreign estimated breeding values had to be converted into Belgian national trait. Also, the associated reliabilities had to be estimated. Then, this prior information was integrated into the Belgian evaluation using a Bayesian approach. It led to a slight modification of the average estimated breeding value and of the standard deviation for the whole population. It also led to a new Belgian ranking of the foreign stallions more similar to foreign rankings. However, despite these modifications, the adequacy of the Belgian model was not damaged. Within regards to prediction ability, Bayesian evaluations using the conversion equation estimated by the Weighted Least Squares procedure predicted the best traditional estimated breeding values of the year 2009 for the French stallions. For the Dutch ones, it were the evaluations associated to the Wilmink‟s conversion equation. For both countries, Bayesian evaluations using Goddard‟s conversion method improved most the stability of breeding values. Finally, integration of French and Dutch information improved reliabilities of the Bayesian estimated breeding values. [less ▲]