Editor's note: The following information is based upon an article written by Bennet Cassell, professor emeritus at Virginia Tech and George Wiggans, research geneticist at USDA-AIPL, entitled, "Adjustments make all cow PTAs similar to those on genomically tested cows." Their complete article may be found at www.holsteinusa.com  Some of the tables and information below have been taken directly from the referenced article.

Background – Why More Rescaling?
In April 2010, USDA introduced two new procedures to the national genetic evaluations, which generated a great deal of discussion and interest within the industry.

First, imputation of genotypes, cows with a number of genomically tested progeny (typically five or more) had their genotypes imputed. Meaning, if a cow had enough information available between her genomic-tested offspring and sire to be able to derive (or "impute") her genotype with a high level of accuracy, she received a genomic prediction.

The second change, which had a larger impact on cow PTAs, was USDA "rescaling" PTAs of cows with genomic information. Rescaling is the process of scaling down the difference between the cows' PTAs and her PAs (Parent Averages). The rescaling of PTAs, in 2010, was done only to animals with genomic predictions, which created the important problem that, since non-genomic animals had not had their PTAs rescaled, the PTAs of genomic-tested cattle could no longer be compared side-by-side with PTAs of non-genomic tested cattle.

What Will Happen in April 2011
Beginning with the April 2011 genetic evaluations, USDA-AIPL will adjust the PTAs on all Holstein, Jersey and Brown Swiss cows, whether genomic tested or not, to make all evaluations more directly comparable. It is important to note that ONLY PTAs for milk, fat (pounds and percent) and protein (pounds and percent) are affected. The improvements made to the PTAs on older cows will trickle down to the genetic evaluations of their progeny. This process will result in better estimates of parent averages.

Rescaling of cow PTAs begins with the within-year Parent Average. To do this, cows are broken down into 10 groups (for each birth year) based on their PA. PTAs on cows with a low PA are scaled up and PTAs on cows with a high PTA are scaled down. If the difference between a cow's PA and the average PA for the birth year is large, larger adjustments will be made. The correlation between the current PTA Milk and the adjusted PTA Milk is over 0.99, so changes in rank are expected to be small for almost all cows.

The main factor in determining the all-cow adjustment is how much an animal's PTA differs from its Parent Average. As a generalization:

·        If the PTA is greater than the PA, the adjustment will cause the PTA to drop
·        If the PTA equals the PA, then there is almost no adjustment
·        If the PTA is less than the PA, the adjustment will cause the PTA to increase

Research has shown the adjusted PTAs of cows included in the genomic evaluations still tend to be slightly biased upwards. Therefore, the traditional PTAs of any Holstein cow that obtains genomic information will be subjected to a reduction of 221 pounds of milk, 8.8 pounds of fat and 6.4 pounds of protein. This represents the average difference between the overall combined PTA on genotyped cows and the PTA obtained directly from the genetic markers. This adjustment is much smaller than the one made in April 2010.

Example: PTA Milk

The rescaling occurs in three steps:
A.   Within year of birth, adjustment for PA.
B.   Adjustment for difference between PTA and PA.
C.   Average adjustment for all genomic tested cows.

A. Within year of birth, adjustment for PA for cows born in 2005
The PTAs on high families are brought down and the PTAs on low families are brought up.

                Low Parent Average Cows                     >>>>                     High Parent Average Cows

Group

1

2

3

4

5

6

7

8

9

10

PA Milk

-834

-383

-202

-71

+46

+158

+274

+405

+577

+924

Change in PTA

+138

+75

+46

+25

+8

-9

-27

-47

-76

-131

New Adj. PTA Milk

-751

-350

-184

-59

46

151

259

381

542

865

B. Adjustment for difference between PTA and PA.
Rescaling of individuals, within a family, also takes place. Individual cows, where their PTA is much greater than their PA, are brought down. And for cows where their PTAs are lower than their PA, their PTAs are brought up. The actual amount of change for each individual adjustment will also depend upon her reliability value. Cows with a high reliability will have less of an adjustment; cows with a low reliability will have more of an adjustment.

C.   Average adjustment for all genomic tested cows.
Newly genomic tested cows will receive a third reduction. Another 221 pounds of milk, 8.8 pounds of fat and 6.4 pounds of protein is subtracted from the traditional PTAs of Holstein cows before being included in the genomic prediction population. This helps ensure the accuracy of all genomic tested animals.

End result:

The new “all-cows” adjustment will have a minor effect on cows which previously received a genomic PTA. The average PTA Milk for the top 100 cows with genomic information will change by -24 pounds. This is because these cows were already adjusted last April. For them, the new procedure is just a refinement. The top 100 cows without genomic information will see a much larger change, -397 pounds, similar to what happened to the genomic cows in April 2010.

PTA Milk for Top 100 Holsteins

Mean (lbs)

Standard Deviation

Minimum

Maximum

         Genotyped

December 2010

1783

368

1101

2719

April 2011

1759

420

825

2555

       Non-genotyped

December 2010

2268

396

1175

3105

April 2011

1871

328

993

2494

The difference in PTAs for cows with genomic information versus those cows without is much smaller. In April 2010, the top 100 cows without genomic information were +485 higher. In April 2011, the difference will be +112 pounds.

One would expect that the top 100 genomic tested cows would be superior. And this will be the case in April 2011. Remember, a top 100 cow without genomic information will experience a 221 decrease in PTA Milk, when they are genomic tested. Therefore, if both groups are genomic tested, the current top 100 genomic-tested cows remain at 1871 and the top 100 non-genomic cows would experience a 221 reduction in their PTA Milk.

       April 2011

Mean (lbs)

Standard Deviation

Minimum

Maximum

Genotyped cows

1759

420

825

2555

PTA of Non-genotyped  cows after they’re genomic tested

 
1650

328

772

2273

In the end, this rescaling will improve the overall accuracy of genetic evaluations, and will be a more accurate predictor of true genetic merit. Because the adjusted PTAs will impact a cow's progeny, it is expected that PAs will be a better indicator of expected merit. PTAs of genotyped and non-genotyped cows will be able to be compared more fairly and accurately than what was possible in 2010, and the overall result of the changes will be an improved method of identifying outstanding females.

Click here to see the attachment to this message for the complete article by Cassell and Wiggans. For more information about the April 2011 Holstein genetic evaluations, visit www.holsteinusa.com.

Source: Holstein Association USA Staff