FAQFAQ      RegisterRegisterLog inLog in



 
        BREEDING FORUM >> Sales Select Report
Author Message
brogers



View user's profile

Posted: Fri Feb 03, 2012 15:43 pm

At the request of "Irish Paddy", I will give a brief rundown of what our Sales Select is. I didn't really start posting on this forum to get a 'product review', but here goes. Smile.....

One thing that needs to be understood with the current state of play in regards to prediction of racing performance on untried stock. At this point, anyone who tells you that they can determine with accuracy the optimal distance and racing class of a racehorse with a single test (genetic or otherwise) is leading you astray. I bolded that statement because I think it is probably the most important thing to understand and I hope that at the end of this you will see how/why.

Firstly, we took a holistic view in terms of trying to evaluate performance. There is not just one single test that can solve the riddle, and even the tests that we do still results in horses becoming stakes winners when the model says they should not. It is prediction, and there are always exceptions to the rule.

To start, we used the recently released 70KSNP chip. Im not sure what everyone knows about genetics, but this chip basically has a set of markers (~65,000) that are placed along the 22,000+ genes of the equine genome. What this allows you then to do is compare two groups and see what variations within genes occur in one group and not another. So for example you could get a group of horses that are wobblers, and a group that are not, and if the trait or disease is something that is occurring because of a genetic change in the DNA code then this will show up. Rather than looking at a disease, we took performance. We gathered 400 samples of horses, 200 horses that were G1 winners, and 200 that were horses that were well bred (so for every good horse by say Distorted Humor we got a slow one also), but had failed to run a 78 or better Beyer speed figure in at least 5 starts and had no known disease or injury. Of the groups of 200 we split them down so that they were even between fillies and colts and then also between sprinters and distance horses, so for example, there were 50 elite colt sprinters and 50 slow fillies that ran a distance.

From there we just compared each of the groups. We started out looking at all sexes, fast vs. slow, slow sprint colts vs. slow sprint fillies, fast sprint fillies vs, fast sprint colts, fast sprint colts vs. fast route/distance colts, etc. What we ended up finding was there were about 200 variations within genes (where elite horses might have had a G:G variant and non elite have an a:G variant or an A:A variant) that made elite horses what they are and the non-elites what they are. The variants are scattered across all chromosomes, including the X (for those that believe in the X-Factor, the concept that a variant within a gene on the X confers a large heart, sorry, that is a myth). What we then did is look at them and find the ones that made the largest difference in terms of the population, and put them in a prediction model that uses a weighted genomic model to predict both on distance (0 being an absolute sprinter and 100 being an absolute distance horse) and class (0 being slow and 100 being fast - in terms of class). Once we developed the model we then went and tested another ~300 horses. Most of these were done blind and the model was quite predictive. Was it 100% accurate - no. It got some horses wrong, including a couple of really good G1 winners, but it did, in a statistically significant fashion, classify the elite horses as elite and the non-elite horses as non elite. There is an interesting factor with distance that we found. Horses that were genetically tested to be sprinters, could outrun this from time to time. As I posted in another thread, as late two-year-olds and early three-year-olds, against their age group, horses that are sprinters can run a little father than they should, especially when the pace is average.

So why did we not just do genetics alone, well, the problem with just doing baseline DNA is that it misses translational and epigenetic influences (environmental interaction with genes) and it also doesn't account for what happens at the metabolic level. Also, no matter how many SNPs you use, you cannot account for all genetic variation. So, we took the view that it would be better to create a prediction model based on a clinic-genomic outcome. Rather than just looking at DNA, we also added in a cardiovascular and splenic evaluation to get a more rounded view of what the capacity of the horse actually is.

Maximal oxygen consumption has been shown in peer-reviewed clinical studies to be a good predictor of equine athletic performance. The cardiovascular system forms part of the oxygen transport system and thus is a key determinant of performance. To put it quite simply, If you require less of your Vo2max to perform in the given level of competition, you will be less fatigued and able to run faster/farther. Resting ECG (heart) measurements have been used to identify elite performers as measurements of a horse’s heart at rest have been correlated to the same horse’s cardiac function during maximal exercise. The predictive value of ECG is similar at each measurement time as long as similar criteria are used to obtain the measurement. In addition to cardiovascular function, the role of the spleen is also important and is measured. The spleen acts as a cardiovascular reserve modifying the heart rate response to submaximal exercise by increasing the circulating red cell volume. This splenic reserve may assist ventricular filling at high heart rates making this an important factor in elite performance.

That all said what we have found is that there are sometimes horses with great genetic profiles, that have very poor or average cardio capacity. Sometimes these are horses that were quite sick as foals so an environmental challenge has in some cases rendered whatever their genetic profile is as rather meaningless (there has been some work done on athletic signatures in blood metabolites which looks really interesting, but we doubt that we will ever be able to draw blood at a yearling sale to test for it!). Sometimes the genetic profile is just confounding data with the cardio and splenic data. Putting this all together was a challenge but a combination of established statistical, phenotypic and genomic covariates leads to better prediction of the outcome than what can be obtained from any one field. Multivariate discriminant regression models, which Performance Genetics has used to generate their prediction model, are very useful in discriminating horses into groups, so elite sprinters versus non-elite sprinters, and the programs are capable of conceptualizing a large number of variables at once to see how they all relate.

So are we happy with what Sales Select represents? Absolutely. We think it is a very solid prediction model. But like all prediction models, there are some areas that are more predictive than others. One area that we feel that we have well worked out is elite sprinters, especially colts. When a horse generates a low distance score (below 40) and a high class score (above 80) and has a certain cardio and splenic type, we are very confident about its chances as a racehorse, all other things being acceptable (reasonable type, passes the vet/scope, etc). There are not a lot of horses that fall into that 'area' and the ones that have in our model are really good. Distance fillies aren't quite as predictable. Is this the end of our research? - no. We are looking towards mitochondria, examining fat-free mass, fascicle length and pennation angle of muscle, and a biomechanical model for prediction in order to get an even more rounded view of prediction of performance.

I hope you haven't gone to sleep by now.... Very Happy
Reply
Back to top
Jump to: