BREEDPLAN changes use of imported EPDs

Introduction

When Version 4.1 of BREEDPLAN was introduced in 1998, one of the improvements it contained was the ability to utilize overseas EPDs in calculating EBVs for animals introduced from overseas. Before this development, all new animals started at zero until data was collected on their performance in Australia. The new improvement allowed overseas animals to receive starting EBVs based on their performance in the country of origin. In setting up this procedure, BREEDPLAN aimed to have starting EBVs based on imported information and then, as Australian data was collected, the emphasis would shift from the imported EPD information to our locally collected performance information.

For the last two years, concerns were raised from researchers and breeders that the overseas information being over emphasized at the expense of the local information. AGBU researched this issue and confirmed that there were some anomalies in the use of overseas information in calculating Australian EBVs.

Findings
Investigation of the procedure and data has revealed that some overseas animals were receiving more benefit from their overseas EPD than was first intended. This was mainly due to the pedigree structure of the immigrant data that had developed over time (see tables below). For example the immigrant EPD file for Angus now contains large numbers of progeny for bulls like USA036 and USA1680 with EPDs that were only pedigree estimates. The old procedure tended to double count this information. This had to be corrected. Also the genetic evaluations of overseas countries has expanded beyond growth traits however, unlike BREEDPLAN, they are often derived from a number of separate evaluations.

Summary
Therefore “importing” was reviewed and alterations made to better reflect the intentions of the process designed to use the overseas information as a “starting value” for the BREEDPLAN EBV until sufficient data has been collected to adequately reflect the immigrant individual’s performance locally.

Several full BREEDPLAN runs were performed to investigate the effect of different procedures and calculate the correlations for the traits between the different runs.

Some important observations:

• The correlation between the old and new EBVs across the traits ranged from 0.96 to 0.99 indicating that across the full analysis, little has changed.
• Some individual bulls, particularly those with extreme US EPDs for a trait or with large numbers of progeny in the immigrant file, have experienced some changes in EBV (up and down).
• Some immigrant animals with extreme EPDs, particularly for mature cow weight and carcase traits, may also have changed in the latest BREEDPLAN run due to the level of emphasis put on the imported information and the way Australia handles correlated or related information.
• The correlations between the US EPDs and the new and old BREEDPLAN EBVs have remained about the same for the growth traits (except mature cow weight) but have reduced for the carcase traits with the new EBVs. This shows the new method is putting less emphasis on the foreign carcase data than the old method.

Why the EBVs of some high accuracy overseas AI sires have changed.

The intent of importing EPD information was to give O/S animals an initial starting value in Breedplan. The impact / effect of the EPD on the imported animal's local EBV should diminish as domestic performance information on the animal’s progeny is submitted to Breedplan. In cases where semen of many sons of O/S bulls was also imported and used, this did not always happen.

In Figure 1 the imported Carcase Marbling EPDs (Fall 2004) for some of the prominent US sires are indicated relative to the Carcase Marbling EPDs of the other animals (about 1500) 1that were imported for the latest Breedplan analysis.

Figure 1: Prominent US sires compared to all the imported AAA Carcase Marbling EPDs.

B/R New Design 036

Question:

Why did 036 change so much for IMF? With a 99% accuracy one would not have expected him to change so much.

Answer:

The change in his IMF EBV had nothing to do with the accuracy. It happened because the whole method of handling imported EPDs and emphasis put on those EPDs changed, not because more data was included.

The calculation of accuracy is a function of:
1) the number of progeny measured for that trait in effective contemporary groups,
2) the correlation with other traits and
3) the heritability of the trait.

In this case the change in IMF EBV wasn't the result of additional data analysed in the same way. It was purely due to the changed emphasis put on imported EPDs. 036 had many of his son's imported and used through AI in Australia. With the old procedure these son's EPDs also contributed to 036's EBVs and therefore over-emphasized some of his EBVs. His carcase marbling EPD that was imported for the January 2005 analysis ranks in the top 5% of US sires.

Local carcase information.
Lets have a look at how progeny of 036 compared locally for carcase IMF.

In contemporary groups including 036, there were a total of 364 animals measured in contemporary groups representing 46 progeny of 036. The mean for IMF% of all 364 animals was 5.48. The mean for 036 progeny was 6.2 and for the other 318 animals it was 5.37.

This indicates that 036 progeny is performing above average for Carcase IMF.

Local scan information.
Scan IMF information is correlated with carcase IMF and is therefore used in the analysis to help with the estimation of carcase IMF EBVs. Figure 2 indicates the frequency distribution for animals scanned for IMF in contemporary groups representing some 036 progeny.

Figure 2: IMF Scan information of animals in contemporary groups containing USA036 progeny.

There were a total of 16986 animals represented in contemporary groups that contained some 036 progeny.
The mean for the 4313 progeny of 036 was 4.02.
The mean for the 792 progeny of USA 5321 was 4.75.
The mean for the 274 progeny of USA 1531 was 3.77.
The mean for the 299 progeny of USA 323 was 3.88.
The mean for the 90 progeny of USA 1407 was 4.20.
The mean for all animals excluding 036 progeny was 3.84.
The mean for all the animals was 3.89.

In this case, both local carcase and scan data supports the imported EPD information indicating that USA 036 progeny is performing well for carcase IMF. He is currently in the top 5% for IMF in the breed in the USA as well as in Australia even though his EBV has decreased to reflect the Australian data.

C A Future Direction 5321

Question:

Future Direction is also a prominent, high marbling O/S AI bull. Why didn't his IMF EBV change?

Answer:

The carcase marbling EPD that was imported and used for Future Direction isn't very high and he therefore did not receive very much benefit from his EPD in the past. His local scan progeny also performed very well when compared with the mean of the contemporaries that they measured against. This helped to sustain his high IMF EBV. This is a case where the local scan information strongly supported the imported EPD information.

Figure 3 shows the distribution of local IMF scan information for animals scanned with Future Direction progeny.

There were a total of 19256 animals represented in contemporary groups that contained some Future Direction progeny.
The mean for the 783 progeny of USA 036 was 3.69
The mean for the 3811 progeny of USA 5321 was 4.44.
The mean for the 390 progeny of USA 1531 was 3.70
The mean for the 103 progeny of USA 323 was 2.91
The mean for the 14 progeny of USA 1407 was 4.71.
The mean for 15445 animals excluding Future Direction progeny was 3.84.
The mean for all the animals was 3.96.

USA 5321 had progeny performing well above contemporaries. Hence his EBV is high.

Rockn D Ambush 1531

Question:

Why did Rockn D Ambush change for IMF?

Answer:

The carcase marbling EPD that was imported and used for Rockn D Ambush was in the top 5% for AAA therefore he received some benefit from his EPD in the past. His local scan progeny didn't performed very well when compared with the mean of the contemporaries that they were measured against and this partially contradicted the US EPD information.

Figure 4 shows the distribution of local IMF scan information for animals scanned with Rockn D Ambush progeny.

There were a total of 6434 animals represented in contemporary groups that contained some Rockn D Ambush progeny.
The mean for the 308 progeny of USA 036 was 3.75
The mean for the 477 progeny of USA 5321 was 4.72
The mean for the 1107 progeny of USA 1531 was 3.54
The mean for the 2 progeny of USA 323 was 3.25
The mean for the 16 progeny of USA 1407 was 5.71.
The mean for animals excluding Rockn D Ambush progeny was 3.86.
The mean for all the animals was 3.80

Rockn D Ambush figures dropped because his progeny did not perform as was predicted by his US information.

B/R New Design 323

Question:

Why did B/R New Design 323 change for IMF?

Answer:

The carcase marbling EPD that was imported and used for B/R New Design 323 was in the top 1% for AAA therefore he received quite a bit of benefit from his EPD in the past. He also had some progeny used as AI sires in Australia through which he received some additional benefit. His local scan progeny didn't performed very well when compared with the mean of the contemporaries that they were measured in and this contradicted the EPD information.

Figure 5 shows the distribution of local IMF scan information for animals scanned with B/R New Design 323 progeny.

There were a total of 6030 animals represented in contemporary groups that contained some B/R New Design 323 progeny.
The mean for the 487 progeny of USA 036 was 4.24
The mean for the 112 progeny of USA 5321 was 3.79
The mean for the 4 progeny of USA 1531 was 4.08
The mean for the 951 progeny of USA 323 was 3.72
There were no direct comparisons with USA 1407
The mean for animals excluding B/R New Design 323 progeny was 3.59
The mean for all the animals was 3.61

Bon View New Design 1407

Question:

Why did Bon View New Design 1407 change for IMF?

Answer:

The carcase marbling EPD that was imported and used for Bon View New Design 1407 was in the top 5% for AAA therefore he received some benefit from his EPD in the past. He is also a B/R New Design 036 son that would suffer due to the change in 036's IMF EBV. His local scan progeny did perform reasonably well when compared with the mean of the contemporaries that they were measured in and this supported the EPD information.

Figure 6 shows the distribution of local IMF scan information for animals scanned with Bon View New Design 1407 progeny.

There were a total of 1911 animals represented in contemporary groups that contained some Bon View New Design 1407 progeny.
The mean for the 41 progeny of USA 036 was 3.92
The mean for the 11 progeny of USA 5321 was 5.00
The mean for the 9 progeny of USA 1531 was 4.17
There were no direct progeny comparisons with USA 323
The mean for the 474 progeny of USA 1407 was 4.17
The mean for animals excluding Bon View New Design 1407 progeny was 3.79
The mean for all the animals was 3.87

From all the above comparisons it is clear that progeny of Future Direction performed very well in nearly all contemporaries which is in line with his high IMF EBV. Some of the other bulls performed reasonably well in some cases (eg. 1407 with 16 progeny when compared with 1531) but overall non performed as well as Future Direction.

Future imported animals with high US EPDs will initially be treated more conservatively in calculating EBVs until they have Australian progeny records. The EPDs are still used but they will not be weighted as highly as in the past and once they receive Australian information, the local information will receive increasingly higher weighting.