GENOMIC EVALUATIONS-- AN OVERVIEW

A visit with Dr David Selner -- January 17, 2009

The January 2009 USA Sire summary release coincided with the first publishing of Genomic (DNA mapped) trait evaluations, as released by AIPL, Holstein USA, AJCA, etc.    This was the result of a collaborative effort between USDA, seven major AI systems (includingTaurus-Service, Inc), breed associations, universities, and other industry entities concerned with genetic advancement.

"Genetics" as a field of human study began with the curiosity of Gregor Mendel, a Cistercian monk from the 13th century, who discovered basic genetic action from breeding peas in the monastery gardens, and wrote down all his observations.      Lord Bakewell, an English landowner with large cattle herds, quantified this further in the 18th century with breeding theories that were influential in the development of many English cattle breeds (Hereford, Angus, Shorthorn, Ayrshire, Jersey, etc).  It is from Bakewell and his contemporaries we developed concepts like linebreeding, outcrossing, inbreeding, etc.

Within USDA's Bureau of Animal Industry, research herds developed through the 1940s- 1950s which were used in breed trait comparisons, crossbreeding research, and in the formulation of the first sire summaries-- published as "herdmate" comparisons since 1964 by Animal Improvement Programs Laboratory (AIPL) at Beltsville, MD. 

Up to the current decade, the word "genetics" in the dairy industry usually meant "populaiton genetics", which is a more statistical than biological evaluaiton and summarization of traits known to be heritable and easily measured for pedigree relationships.      But with the advent of "gene mapping" (done first on mice, then humans, now bovines) we now have "Genome" information, ie, the ability to relate traits observed to physical genes possessed.

The first "Genomic" measurements were done by the Swiss, who were intensely interested in the genetics of Cheese yields, and identified the Kappa Casein gene variants (A and B), verifying that possession of the "B" K/C gene results in 10% higher curd formation (cheese yield) from a given protein and butterfat content milk, than possession of the "A" K/C gene.  This research found that 92% of Brown Swiss, 78% of Jersey, but only 19% of Holstein cattle carry the "B" variant for cheese.

The next Genomic breakthrough was identifying the genes indicating possession of various lethal recessive traits-- Pinktooth and Mulefoot in Holsteins, Limber Leg and RVC in Jerseys, Weaver in Brown Swiss.    (Later, DUMPS, BLAD, CVM were also found through DNA analysis.)     This was a great help in allowing genetic selection for AI sampling and ET production among descendants of known carriers of lethal gene recessives.      DNA tests were then developed or are developing for the more desirable dominant and recessive traits-- homozygous polled, red hair color, antibody caseins, etc. 

 [continued]

CHROMOSOMES AND GENES

In Bovines, we find 30 chromosomes.     On each chromosome are found gene pairs, which are made up from four amino acids: A, C, G and T.     The sequencing of these amino acids appears to determine the qualities of each gene.     (Note that 40% of the genes possessed by bovines are "in common" with genes possessed by humans.)   The sum total of all these genes for each specie is call the "Genome" for that specie.

The prior mapping of mice, human and canine Genomes actually assisted in developing the Bovine genomic identifications.   For instance, the BLAD recessive has been found in dairy cattle, Irish Setter dogs, and humans.     All mammals likely share some genes in common.

The measurement of trait action and its relationship to the Genome has been accomplished by a separation of DNA strands into "Single Nuclear Polymorhic strips" [SNPs].    This process is done by the chemical addition of enzymes that separate the DNA strands into linked genes.    Four bio-tech companies have bene involved in doing this part of the research process for AIPL.    The results of 50,000+ strips are being measured by AIPL computers against animals of known 99% Reliable trait data, to create the evaluaiton "map" of the bovine genome.

Other countries have pursued Genomic testing-- LIC New Zealand, for example, is now two years into the use of Genomic estimates for sire evaluations, and utilizes a similar number of SNPs as the USA effort.     (Oddly, New Zealand has found Jersey G data more accurate than Holstein G data to date.)     Denmark's DNA testing was using 1500 SNPs to focus on individual traits they deemed important.

Apparent variation among traits being measured

70% of the "milk yield" value of a bovine is coming from the DGAT gene, which is unusual compared to the trait impact of most genes.     Genes influencing production traits have been found on 25 of the 30 chromosomes.    By comparison, conformation traits appear mostly on three chromosomes; health traits appear mostly on four chromosomes.

     Some genes are "interactive", ie, are more dependent on a mating effect;  some genes express an "over-dominance" producing a result you can describe as similar to heterosis resoponse.     There may also be "accelerator" genes that act as a stimulant to the action of other genes.     Thus far, Genomic testing is unable to measure such effects, so this contributes to our inability to assume a Genome "proof" can be as individually accurate as a progeny sampling evaluation.

But to this point, the accuracy of Genomic testing has led AIPL to assign twice the Rel% to the current Genomic estimates as it assigns to Parent Averages-- a range of 60% to 70% Rel, compared to 35% to 40% Rel for ancestry averaging, for each measured trait and composite index. 

Dr Selner's initial conclusions

David Selner, Ph D genetics (formerly staff geneticist for Genex and Alta genetics), has read all the major literature on the subject and after reviewing the first published summary data, drew these conclusions:

(1)   Over all animals tested to date, the average of G PTAs is slightly lower than the average of PA PTAs, suggesting (as was hoped) genomic testing would help eliminate some random inaccuracies from progeny testing evaluations and cow indexing estimations.   The previous "top 20 TPI" sires dropped an average of 100 points after adding Genomic data.

(2)   The very highest Genomic tested animals have high levels of pedigree relationships.   This would suggest that utilizing G PTAs without waiting for progeny verification could lead to much higher inbreeding levels.    This could be the fault of the highly selective pedigrees initially chosen for genomic testing: through December 2008, only 138 of the top 5000 CTPI Holstein cows had been genomically tested.

(3)   The Rel% of individual traits tested by Genomics is as high as 80%.    The lower the heritability of the trait in measurement, the more variable appears the Genome pattern in animals measured for that trait.      Thus Genomics is not a "replacement" for other breeding techniques (such as aAa), but is more a refinement of our ability to read genetic potential beyond knowing the pedigree of the animal.  

(4)   While Genomics adds accuracy to the evaluation of tested animals, it is not likely to increase the accuracy of the PA estimation of the progeny of two G tested parents.    Biology's random pairing of genes at conception remains the operative factor.    (Full siblings are already known from G testing to vary by as much as 40% in major trait areas.)

(5)   Full siblings with mapped Genomes can now be sorted to individual levels of inbreeding, ie, the actual numbers of homozyogus gene pairings are known.    Our thinking about the issues around inbreeding may change, givne this ability to further compare full sibling evaluaitons against their Genomic trait estimates.

(6)   Individual anomalies in the G results will lead the afflicted AI systems to question the first G evaluations' collective accuracy.    For example, there is a Holstein sire with 50,000 tested progeny in eight countries, rated +3.0 PL in August 2008, whose G value for PL was much closer to 0.0.   One of the major participants experienced a statistically significant drop in PTAs for their entire active lineup.     This particular AI system has a highly restrictive herd size and herd average qualification to be involved in their YSP progeny testing, so it may indicate that such sampling restrictions do result in over-evaluation...    It will take time to sort out these aberrations.

Dr Selner's recommendations

Genomics appears to be a powerful new tool.    How it will be individually applied by industry participants will be crucial to whether its next benefit is a "net" positive or negative.

A major motivation for G testing is to reduce costs of accurate sire sampling.   One major AI system has decided to reduce in half the number of herds targeted for sire sampling-- thus where they used to want progeny in fifty herds, now they will release bulls on the results of a G test and twenty five herds (which will be roughly equal levels of Rel% under current assumptions).

This tells us that a G test (assigned 65% to 70% Rel) is the rough equivalent of 15 daughters in ten herds--  so adding another 25 herds with progeny generates the 85% Rel that exceeds the minimums set for official TPI rankings (and by most semen buyers).    Given the way the "Animal Model" builds a cow and sire evaluation, a G test will replace Parent Average, on the tested individual, but will remain a permanent part of the data set for any evaluation on (or including) that individual animal.

Many young sires with G tests are now being released by AI systems for general marketing, at prices from $12 to $50 per straw.      Whether you are comfortable with paying $50 to use a bull with 0 daughters in 0 herds evaluated is obviously an individual decision.    But given previous experience, it remains logical to use G tested sires in "groups" to insure the average of the progeny produced lives up to the estimations, that can fail on any individual sire. 

Genomics is capable of indentifying possible "outcross" cow lines that can be sources of superior performance ability, exceeding PTA estimates based in Parent Averages.    The problem is the current lack of motivation to test animals who are not already putting bulls into AI, given the cost (generally around $200 for individual females).    But the ability of the highest G ranked sires to be marketed in quantity without progeny verification of their "value" suggests inbreeding could now become the real problem we only theorized earlier.     So breeders and interested AI studs need to find and test the best "outcross" cows and look for the outliers that can add some needed trait emphasis and additional pedigree variety to the market.