What have we learned after six years of genomic proofs
Extended comments are highlighted in blue.
The first dairy cattle genomic proofs were released in 2009. Some six years on, daughters of these bulls have come—and many have already gone. What have we learned?
Dairy Today visited with three University dairy geneticists to find out. Included in the roundtable are Bennet Cassell, professor emeritus of dairy cattle genetics with Virginia Tech, Les Hansen, a dairy cattle geneticist (and crossbreeding advocate) with the University of Minnesota, and Kent Weigel, a dairy geneticists and chair of the department of Animal Science at the University of Wisconsin.
Have the promised gains in reliability (REL) held up? Milk, protein, fat?
Cassell: If by gains in REL, you mean increasing accuracy as the technology matures, I do think that is the case. However, increases in REL due to better chips, more data, better techniques have been relatively minor. The major gains due to knowledge of genomic sequences in young animals seem to be of the same magnitude as they were originally.
Hansen: Yes, reliability increases over parent average from genomic information. However, unfortunately, many people, including those working in promotion and the dairy media, have commented that genomics has provides a “doubling of genetic knowledge” over parent average. That isn’t true. One must mathematically take the square root of reliability to arrive at “accuracy”. Accuracy is the measure of the confidence interval ( or ‒ ) around a PTA. Reporting of reliability instead of accuracy for PTA greatly exaggerates the gain achieved from genomics over parent average.
Also, those individuals (male or female) ranked highest on genomic evaluations are more likely to drop for PTA with more information than individuals ranked more lowly. Finally, genomic PTA are more likely to be over-evaluated (biased upward) if sires of the individual have genomic-only PTA without daughter information.
Weigel: Generally speaking, yes, but there is still some overestimation as compared with future daughter performance. Latest estimates are about 175 pounds of milk, 3 pounds of fat, and 2 pounds of protein per lactation, on average.
What about productive life?
Cassell: This trait will always be challenging since true expression (age when no longer productive) requires time for expression. It seems to me that what has happened is a consistent effort by USDA/AIPL staff to make the estimates of genetic merit for productive life as well as its REL realistic in light of information available at an early age.
Hansen: What applies for production applies for all traits. Productive life is especially difficult to predict even for a bull with daughter information, because cows must be given the opportunity to survive for multiple lactations to accurately predict PL. Genomic predictions of PL are especially impacted if bulls that are close-up in the pedigree are genomic-only or have only very young lactating daughters, which typically is the case.
Weigel: Slightly overestimated as well, roughly 3 weeks, when compared with actual culling data of the daughters.
Have we begun to see more progress in less heritable traits?
Cassell: I think this is happening. We must keep in mind that expression of better genetics for lowly heritable traits is not nearly so clear as for highly heritable traits. In other words, the changes made are often obscured by environmental effects. How do you “see” 2% improvement in genetic merit for conception rate when conception rate of 25-30% is pretty good? And even then, super environmental conditions must be maintained.
Hansen: Genomic predictions are indeed helpful to get an additional “hint” of the genetic worth of individuals for traits with low heritability. Some have commented that genomics “works better” for the lowly heritable traits, but that statement likely is a stretch. Genomic predictions provide improved PTA over parent average for all traits including those with lower heritability.
Weigel: We have made more progress for traits like female fertility, though it’s hard to tell whether this extra project is due to genomic selection, or if it’s because farmers are putting more emphasis on female fertility when making sire selection decisions.
Have we made progress with dairy cattle feed efficiency?
Cassell: Again, we probably have, but it’s tough to see. Today’s dairy cows are much more efficient than cows of my childhood (1950s), but we don’t have any of those old couch potatoes around anymore for comparison.
Hansen: Research continues on the genetic control of feed efficiency (Weigel at UW-Madison is a collaborator in that effort), and the researchers in an international effort will investigate if the potential exists to use genomic predictions for improvement of feed efficiency. A problem for genomic prediction is updated data for traits must be continuously fed into the prediction system, and feed efficiency is heavily influenced by feed intake, which is expensive to collect. The research on genetics of feed efficiency might lead to indicator traits among the current traits collected to improve feed efficiency indirectly via other traits. That would the case with or without genomic prediction.
Weigel: Not yet. Our USDA-funded research project on genomic selection for feed efficiency still has more than a year to go, so we haven’t realized any benefits in practice yet. We’ve compiled feed intake data on nearly 7,000 cows so far, and about 5,000 have genomic test results as well (some cows from historical nutrition studies don’t have DNA available). We should have some preliminary feed efficiency predictions for US dairy bulls within the next year, and then the dairy genetic industry can decide how they’d like to proceed.
Have we been able to more readily eliminate undesirable, even lethal genes from the breeds?
Cassell: Yes. We can detect many of them in animal of young ages and act accordingly.
Hansen: Absolutely, we have been able to detect haplotypes (closely associated with genes) that result in embryo death. The detection of those haplotypes is a very valuable contribution to improvement of fertility within dairy breeds. However, most genetic recessives aren’t lethal and each recessive, individually, will have only modest impact on traits. However, accumulating effects all recessive genes impacting a trait could result in a large impact. Genomics isn’t capable at this point to detect undesirable recessive genes unless they are large and lethal (embryo loss). Some suggest genomics will easily detect ALL genetic recessive traits, but that just isn’t the case.
Weigel: Yes, genomic testing is a fantastic way to find undesirable recessives, and at least a dozen specific genes or haplotypes that are lethal or otherwise detrimental have been located. Total elimination of carriers hasn’t happened yet, nor should it happen, because some of these animals carry other traits that are very desirable. The frequencies of the undesirable alleles can be reduced significantly, and carrier x carrier matings can be avoided quite easily with computerized mating programs.
Do these results mean we can rely less on progeny test programs?
Cassell: We must continue to record phenotypes on genomically tested animals to maintain useful prediction equations. But the use of “progeny test” programs has changed. It is certainly less important than it was prior to genomic testing.
Hansen: Some early proponents of genomics made the assertion that genomics would “replace” progeny testing. However, we have learned that data on cow performance must continue to be collected for genomic prediction to be effective into the future. Therefore, “progeny testing” isn’t eliminated but, rather, is altered in format.
Weigel: Instead of progeny testing, we are now doing “progeny validation.” Semen from young bulls is sold as soon as sufficient quantities can be produced, and because the data collection infrastructure is still in place we eventually get daughter-based predictions of their genetic merit. By this time the bull himself might no longer be in service, and farmers may be using his sons or grandsons. If daughter performance isn’t satisfactory we have to “correct the course”, and that’s why I call it validation rather than testing.
Has the genetic base narrowed and inbreeding increased as we’ve intensively used genomics?
Cassell: I don’t think so. I think genomic testing enables us to increase numbers of progeny from animals of superior merit without having to increase the numbers of progeny from sibs to genetically superior animals which have the same parent(s) but inherited poorer Mendelian gene samples. In this way, we have REDUCED inbreeding. With genomic testing, animals that trace to common ancestors do so through parents, grandparents that were truly genetically superior. In the past, that wasn’t the case, since only about 10% of progeny tested animals (with very similar pedigrees) survived progeny test.
Hansen: By definition, genomics “uses the past to predict the future”. Therefore, only descendants of high-ranking individuals from the past will rank highly with genomic predictions. Besides this, the generation interval is greatly reduced with genomics, and this automatically accelerates the annual increase of inbreeding within a breed.
Accelerated annual increase in inbreeding is an obvious concern about genomics for almost everyone. For the Jersey breed, implementation of genomics coincided with the blending of the totally outcross Danish bull, Impuls, into the U.S. Jersey population. The influence of Impuls relaxed increases of inbreeding within Jerseys for a period of time after the launch of genomics but, over the past year or two, average inbreeding seems to be jumping alarmingly on an annual basis for the Jersey breed.
For U.S. Holsteins, two bulls that were somewhat less related to the breed – Robust and Dorcy – had higher PTA for Net Merit and TPI based on actual daughter performance than on their genomic predictions (because they were somewhat less related to the breed). These two bulls were major exceptions to the rule.
Therefore, at this point, many of the very highest bulls and females for genomic predictions have heavy doses of Robust and Dorcy in their pedigrees. However, moving forward, Robust and Dorcy will no longer be less related to the breed, so average inbreeding within the Holsteins could increase even more annually than the 0.2% we have observed over the past few years.
Weigel: Frankly, the genetic base was already narrowing and inbreeding was increasing prior to genomic selection. We are making faster decisions and using animals (particularly bulls) at a younger age, so that leads to faster genetic progress and faster accumulation of inbreeding.
However, genomic testing allows within-family selection decisions, so the AI stud with first choice will take the best bull calf from among the set of ET (or IVF) full brothers, and the AI stud with second choice will probably move on to another family. In the past, every single full brother went to one of the competing AI studs, because we had no way to choose among them.
Conversely, have we been able to find genetically diverse bulls to slow inbreeding?
Cassell: The AI folks seem comfortable with this statement, but the costs of looking for those outliers always affects the bottom line for breeders and AI. So we may not have searched quite as hard for such animals as we once expected. After you find them, by the way, you still seek to increase gene frequency through heavy use – leading to new opportunities for matings of animals related through common (but diverse) ancestors.
Hansen: No, that isn’t happening. Genomics definitely can detect cattle that are less related to the breed, but the short-term goal with genomics has been to mate the best to the best as fast as we can within the Holstein and Jersey breeds in the U.S.
Weigel: Yes and no. We have been able to find genetically diverse bulls, and lots of “outcross” bulls have been tested. However, their genomic predictions have been poor, and very few of these bulls have actually entered AI service. This is not surprising, because we are taking only a fraction of the top 1% of the population to use as AI sires or ET (or IVF) donors. If we move down the list very far in search of an outcross bull or heifer (based on pedigree value), that animal has a very slim chance of ranking among the breed’s elite after genomic testing.
Other comments/thoughts about genomics?
Cassell: I was really impressed by the response of Amish grazing dairy farmers in the Wooster, OH area to a presentation I made in January on breeding and genomic testing. Those fellows were interested in raising their most promising replacement heifers. Also – Zoetis is a major player in genomics today. That wasn’t the case when I retired in 2010. What might that mean for dairy cattle breeding?
Hansen: For any exciting new technology, the phases of implementation are quite predictable. Immediately after introduction, there is a period of hype and excitement associated with the adoption. Next, comes a period of observation and reflection. Often, this is followed by a period of dis-infatuation and concern for potential unintended consequences. Finally, appropriate adoption of the new technology takes place. We are still working our way through those stages with genomics. Many, many positive outcomes are obvious from genomics, but we are still evaluating the unintended consequences, especially short-term versus long-term consequences on the breeds.
Weigel: This technology has moved from research to implementation at a startling pace, and even today we’re still working out some of the kinks. Genomics does a great job of identifying the top, middle, and bottom groups of animals, but it is far from perfect when it comes to saying which animal is truly #1 and which animal is #2 or #3. The best strategy for dairy farmers is to spread the risk across a larger group of animals. That is, when purchasing semen they should use three times as many young genomic bulls in their herds, as compared with the number of progeny tested bulls they used previously (or swap out the group of service sires three times as often). As for elite cows and heifers, they should recognize that putting all of their eggs into one basket by buying a really expensive calf or spending wads of cash doing IVF work with a single heifer is a risky strategy. They would be better off working with the top 25 or 50% of heifers in their herds and making sure they are mated to the best available AI sires, and potentially with sexed semen. It may not be as exciting, the results will be more predictable.