Minutes from Interbeef Technical Committee Meeting

Aarhus, May 27, 2013
Hotel Scandic Aarhus City

Participants

Brian Wickham (WG Chairman), Lubos Vostry, Pavel Bucek, Zdenka Vesela, Clara Diaz Martin, Eric Venot, Ross Evans, Thierry Pabiou, Raphael Mrode, Anders Fogh, Valentina Palucci, Mohammad Nilforooshan, João Dürr, Eva Hjerpe, Japie van der Westhuizen, Laurent Griffon, Friedrich Reinhardt .

1. Opening and adoption of agenda

The Interbeef Chairman, João Dürr, welcomed all the participants. The agenda proposed contained many reports both from Interbull Centre and the countries. Spain did not receive all the reports sent by countries via email. The chairman reminded the group to use the Interbeef forum so to avoid problems such this in the future.

ADOPTED AGENDA

1. Opening and adoption of agenda
2. Minutes from meeting in Uppsala - Dec 2012
3. Results from the Adjusted Weaning Weight test run
   1. Variance components (Mohammad Nilforooshan)
   2. Breeding values (Mohammad Nilforooshan)
   3. Reliabilities (Mohammad Nilforooshan)
   4. Country reports on results
4. Calving traits pilot run
   1. Data reception (Eva Hjerpe)
   2. Report from CZE (Pavel Bucek)
5. Heterosis and recombination coefficients (Thierry Pabiou)
6. Genetic parameter survey (Ross Evans)
7. Inclusion of new countries in AWW evaluations
8. Timeline for new traits
9. Technical committee terms of reference
10. Recommendations to the working group
11. Adjourn

2. Minutes from meeting in Uppsala

The minutes from the previous technical meeting held in Uppsala in December 2012 were accepted without edits by all participants.

3. Results from AWW test run

a. Variance components b. breeding values prediction c. reliabilities

Mohammad presented an overview of data used at Interbull Centre for variance components estimation, breeding value prediction and reliabilities estimation; the preliminary checks done on the data and the results for correlations, international EBV and reliabilities for all countries and the breeds they participated with. While EBV prediction got a general approval, the main concern from the participants was still on parameter estimation procedure and reliabilities.
Eric Venot pointed out to do not put much attention on the EBV results until we are more confident that variance components estimation is correct. Eric also highlighted the importance to have the standard errors of the estimates available so to have a better understanding of the quality and soundness of the results. Unfortunately MIX99 does not provide such estimates.
Ross Evans commented that the difference options on conversion criteria might have an impact on the final results and wondered if Interbull Centre had tried some different options. Mohammad reported that indeed different scenarios were tested. Some suggestions were outlined:

• Raphael Mrode proposed to validate the results on converging by running a bivariate analysis from two countries;
• Clara Diaz proposed to estimate variance components with a bayesian analysis instead of Mix99, in this way we could see what the probability of a correlation being above a certain threshold would be.

João reminded the participants that estimation of genetic parameters was still an open issue. Interbull Centre implemented an estimation procedure in order to go ahead with the test run and provide updated results, but the complexity of the topic still requires a common effort between the Interbull Centre and ICBF, which is the designated research partner for this topic.

d. Country reports (written reports are attached to these minutes)

SPAIN: EBV results were acceptable but there were some concerns on reliability which values seemed to be too low. Mohammad commented that some differences in international/national reliability were due to the different value of heritability used in the international analysis compared to the heritability used nationally.A plot of national maternal effects against Interbeef EBVs showed no trend, which indicates that in the international evaluations most of the maternal effect is absorbed by the residual and not by the sire effect which is a positive characteristic of the model.

FRANCE: Requested better descriptive statistics about the data structure used for international evaluations.They proposed a change in the distribution file format: having the information displayed vertically rather than horizontally. João informed that a similar change in format is undergoing also in dairy and that therefore it would not be a problem to implement. It was also pointed out a disagreement that pedigree information marked as invalid by the authoritative organization is still included in the analysis. Valentina explained that the reason for it was that animals in such situation were associated with performance records elsewhere and would likely be old animals being exported at very young ages not having their IDs properly recorded in the country of first registration. It was decided to assess the problem to see how many of these animals were included in the analysis, check their birthdate and try to find a way to discriminate these animals from the other validated animals. There seemed to be some issue with reliabilities results as there were no results for the range 0.0-0.08. The reason for this absence was unclear and it was agreed that Interbull Centre will look into it. In general international reliabilities were higher than national reliabilities, a possible explanation for it could be the different value of heritability used and the benefit from using international data.
Eric suggested two ways for validating the results:

1. Interbull Centre should run a new variance components estimation with correlations equal to zero and using the same variances as used by the countries;
2. as the French national evaluation includes also animals not sent to Interbeef, France should run an evaluation with a data set as close as possible to the data sent to Interbull Centre, so to reduce the sources of variation.

IRELAND: as Interbeef evaluation did not include crossbred animals, the results were not yet very useful for Ireland. Inclusion of crossbred animals within a given breed of evaluation would definetely increase the importance of such evaluation in Ireland.

DENMARK: results for Charolais seem reasonable as the trend of IEBV is quite similar with the national trend, the trend for Limousin presents some differences, national trend appear to be much steeper than international trend.

CZECH REPUBBLIC: The results looked much better than the ones evaluated in December. There still seems to be some issues in the genetic parameter estimation.

UNITED KINGDOM: There seems to be an issue in the scale of expression for IEBV. Last results looked better than the ones presented in December, and there was a significant amount of reranking.

SWEDEN: The report sent by email showed to be in line with the other countries' report.

Final remarks: International EBV estimation using Mix99 have proved to work well, majority of trends were good although for some countries there could be a scale issue to investigate. For genetic parameters and reliabilities estimation Mix99 might not be the best choice. It was suggested to estimate genetic parameters and reliability using a small dataset and running with both Mix99 and a second software (either DMU or ASREML) and compare results. DMU would be a better choice as it could also provide standard errors of the estimates. Having a comparison with another software we could ask MTT to inspect better the differences and provide advice on how to proceed.

4. Calving traits

a. Data Reception

Eva presented an overview of the data received from the participating countries on calving traits, the most common errors in preparing the necessary files and the amount of data sent to Czech Republic for analysis.

As all the information of pedigree have been recoded prior sending them to CZE, it was agreed that for better understanding and presenting the results the Czech could get the full international ID of all their national animals plus BREED+COUNTRY for every other international ID.

b. Report from CZE

Pavel presented a general introduction of the data received. There research group was waiting for some extra information coming from some countries in order to continue the research.

5. Heterosis and recombination

Thierry Pabiou presented the results on the possibility to calculate recombination and heterosis starting from the Interbeef international pedigree. In his study he compared the results between the ICBF and Interbeef databases. The main point was the availability in ICBF of additional information in case of missing pedigree, i.e. availability of the breed of the sire/dam, which helped in the calculation of breed composition and therefore in the recombination results comparing to the Interbeef database where such information are at the moment not available. Thierry's study pointed out that breed composition and recombination are therefore highly affected by the source of additional information and therefore change drastically between Interbeef and ICBF. One of the many questions raised was what should be collected and make available in the Interbeef database. Some suggested to store breed fraction of founders and breed of missing parents. Friedrich Reinhardt commented the risk to have such information recorded in a permanent database such as the one stored at Interbeef. As the pedigree is a permanent one, as soon as a pedigree from a founder is provided or corrected this will change the breed composition of all other animals linked via the pedigree to this founder and therefore the breed composition will change overtime. Another question to consider when crossbred data is included in the Interbeef purebred evaluations is which is the minimum fraction of the breed of evaluation that is required in order to include the animal in the analyses.

6. Genetic parameter survey

Ross Evans presented a survey on genetic parameter estimation as carry out in most countries. Interbull Centre is still open for collaboration and is still available for running analysis if it is needed but the research development is up to the ICBF team. Some suggestions were provided on how to try to improve the estimation of variance components:

• Work with a well connected data but looking also at the data structure:having enough contemporary groups level and good numbers of herds among sampled data.
• Limit pedigree information to 5 generations.
• Run analysis with Mix99 and DMU and compare results.

ICBF and will present a research proposal in the Interbeef working group meeting.

7. Inclusion of new countries in AWW evaluations

A new evaluation is scheduled for September 2013. Due to the amount of activities already scheduled at the Interbull Centre, the evaluation in September has to be a routine run, which implies that no genetic parameters will be estimated in September. Due to the fact that Germany is ready to send data and therefore is ready to participate to the September evaluation it would be necessary to re-estimate variance components before September. It was proposed that Germany would send proper file for AWW and pedigree to Interbull Centre as early as July 2013. The Interbull Centre will run variance components in July using the new data from Germany and the old data from the other countries, so that the September run would still be a routine run in which Germany would be able to participate having the variance components being estimated earlier.

8. Timelines for new traits

The participants expressed high interest for carcass traits to be the next new trait to start researching. It was agreed that the United Kingdom and Ireland will present a joint research proposal and a timeline for addressing carcass traits.

9. Technical committee terms of reference

Due to lack of time it was agreed that everybody would read the documentation and send comments/recommendations to the forum.

10. Recommendations to the working group

João summarized the main topics and decision of the meeting which he will present the next day at the Interbeef Working Group meeting:

• Need of detailed description of data and evaluation results from the Interbull Centre to participating countries.
• Mix99 is a well tested software for EBV estimation and results look consistent
• Genetic parameters:
  • Continue research because it impacts ranks and reliabilities

  • ICBFwill lead the research:
    

    • Survey among users
    • Sample data looking at connectedness and data structure
    • Limit number of generations with data
    • Try alternative statistical methods (e.g. principal components)
    • Compare Mix99 with alternative software
• Reliabilities:
  • Compare Mix99 with alternative software (fit one country data only to compare results)
  • Investigate gaps at very low reliability values (0.0-0.08)
  • Check Maternal effects in Mix99
  • Compare sources of information considered both at national and international evaluations.
• Pedigree:
  • Investigate extension of "invalid IDs" usage and propose a strategy to deal with these cases
  • Check how many cases there are in the analysis and how old they are
  • Propose way to distinguish between invalid and unverified animals.
• Calving data:
  • Send COU+BREED combination to CZE for research.
• Breed fraction: continue discussion
• Terms of reference: will be review and discussed online

11. Adjourn

The Chairman thanked all participants and adjoured the meeting.