Stan
Why don't you contact Byron and ask him to supply the data that supports his point of view concerning the "broodmare sire effect". I would caution you that the information is more than likely not free to the general public so be prepared to shell out a few bucks.
DDT
Buckpasser in the X
Moderators: Roguelet, hpkingjr, WaveMaster, Lucy
http://performancegenetics.com/how-it-works/
Genomic Selection
For the past three centuries the Thoroughbred has been bred and selected, in a non-random event, for athletic phenotypes, specifically speed and stamina. While the phenotypic adaptations to exercise have been well described in scientific literature, the genes that underlie such adaptions and in turn athletic performance are only now being investigated. Using the latest 70KSNP chip, which resulted in over 26 million data points, Performance Genetics sequenced the genomes of some 400 horses, both elite racehorses and non-elite racehorses. When comparing the two sets of horses we identified polymorphisms (changes in the DNA) in genes that are directly involved in muscle twitch speed, energy production, muscle oxidative capacity and potential athletic ability. After statisitcal analysis of the relative importance of theses changes, it was clear that there are a number of polymorphisms that appear in one form in the elite racehorse and another in the non-elite. By combining these differences in a polygenic profile (a collection of the variants in a mathematical model), Performance Genetics was able to separate out the elite from the non-elite in subsequent blind tests. While the combination of the polymorphisms can differ between each horse, the elite horses, at both sprint and route distances, had significantly more of the variants associated with superior performance than the non-elite thoroughbred. The important thing to note is that visually, horses with variants that are more found in non-elite racehorses are indistinguishable from their successful counterparts, who have a different sequence which allows them to rise to the elite of the population.
Bring it together – Using Statistical Modeling
As good as each of these fields stands alone in selecting the elite racehorse, a combination of established statistical, phenotypic and genomic covariates leads to better predictions than what can be obtained from any one field. To do this, one uses statistical regression models that take all the variables that each of the entities has and establish what variables have an impact on the outcome. 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 relate. Regression models that perform this function have long been used in medicine, clinical epidemiology, health services research, pharmaceutical research, social sciences and business studies. Their application has greatly increased in the past few decades and regression models can also be used to combine data from different sources. In cancer research, for example, predictive models based on wide range of data such as family history, clinical observations, gene expression and single nucleotide polymorphism (SNP) are frequently developed to assess risk factors with the ultimate aim of developing better prevention, diagnostic and treatment strategies. In recent years, clinical data have been integrated with genomic and other types of data for developing more accurate predictive models For instance, clinico-genomic models were used in breast cancer outcomes prediction where genomic data combined with clinical and demographic characteristics were used for improving predictive accuracy.
The Performance Genetics Model
Performance Genetics is applying a clinico-genomic model in the prediction of elite performance in thoroughbreds. It is taking the best statistical data, cardiovascular measurements and genomic information available and putting these data points into a model that can predict the potential for a horse to become an elite racehorse. Like any model, it is a prediction of the most likely outcome. A racehorse must be fast, but above all, it must be able to adapt to the particular conditions prevailing in each event. This makes the actual performance of one thoroughbred somewhat difficult to define as it is relative to competition and circumstances of this competition that it faces on the given day. Despite this challenge, to make any model work, we firstly had to define what an elite horse and a non-elite horse is. There are a number of ways that we could have done this but in terms of what was reproducible and significant, for the purpose of our preliminary studies our elite horses were Grade One winners that earned a peak Beyer Speed figure of 108 or more. This group is in fact the elite of the elite in many ways and this population included the likes of Azeri, Commentator, Cigar, Caller One, Formal Gold, Serena’s Song and Turkoman. It included 10 of the past 18 winners of the Eclipse Award for Champion Racehorse and many household names so in terms of elite, the initial group of 200 horses was certainly that. The non-elite horses were also selected. These horses, in the main, were well bred horses by the likes of A.P Indy, Distorted Humor, Storm Cat, etc but they were slow. They had to have had at least 5 starts, no known impediments to their level of performance (surgeries, etc) and have not run a peak 78 Beyer Speed figure or better. After defining the model, we had to both train and then test the model. The training of the Performance Genetics model was completed on some 400 horses, 200 elite racehorses and 200 non-elite.
Genomic Selection
For the past three centuries the Thoroughbred has been bred and selected, in a non-random event, for athletic phenotypes, specifically speed and stamina. While the phenotypic adaptations to exercise have been well described in scientific literature, the genes that underlie such adaptions and in turn athletic performance are only now being investigated. Using the latest 70KSNP chip, which resulted in over 26 million data points, Performance Genetics sequenced the genomes of some 400 horses, both elite racehorses and non-elite racehorses. When comparing the two sets of horses we identified polymorphisms (changes in the DNA) in genes that are directly involved in muscle twitch speed, energy production, muscle oxidative capacity and potential athletic ability. After statisitcal analysis of the relative importance of theses changes, it was clear that there are a number of polymorphisms that appear in one form in the elite racehorse and another in the non-elite. By combining these differences in a polygenic profile (a collection of the variants in a mathematical model), Performance Genetics was able to separate out the elite from the non-elite in subsequent blind tests. While the combination of the polymorphisms can differ between each horse, the elite horses, at both sprint and route distances, had significantly more of the variants associated with superior performance than the non-elite thoroughbred. The important thing to note is that visually, horses with variants that are more found in non-elite racehorses are indistinguishable from their successful counterparts, who have a different sequence which allows them to rise to the elite of the population.
Bring it together – Using Statistical Modeling
As good as each of these fields stands alone in selecting the elite racehorse, a combination of established statistical, phenotypic and genomic covariates leads to better predictions than what can be obtained from any one field. To do this, one uses statistical regression models that take all the variables that each of the entities has and establish what variables have an impact on the outcome. 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 relate. Regression models that perform this function have long been used in medicine, clinical epidemiology, health services research, pharmaceutical research, social sciences and business studies. Their application has greatly increased in the past few decades and regression models can also be used to combine data from different sources. In cancer research, for example, predictive models based on wide range of data such as family history, clinical observations, gene expression and single nucleotide polymorphism (SNP) are frequently developed to assess risk factors with the ultimate aim of developing better prevention, diagnostic and treatment strategies. In recent years, clinical data have been integrated with genomic and other types of data for developing more accurate predictive models For instance, clinico-genomic models were used in breast cancer outcomes prediction where genomic data combined with clinical and demographic characteristics were used for improving predictive accuracy.
The Performance Genetics Model
Performance Genetics is applying a clinico-genomic model in the prediction of elite performance in thoroughbreds. It is taking the best statistical data, cardiovascular measurements and genomic information available and putting these data points into a model that can predict the potential for a horse to become an elite racehorse. Like any model, it is a prediction of the most likely outcome. A racehorse must be fast, but above all, it must be able to adapt to the particular conditions prevailing in each event. This makes the actual performance of one thoroughbred somewhat difficult to define as it is relative to competition and circumstances of this competition that it faces on the given day. Despite this challenge, to make any model work, we firstly had to define what an elite horse and a non-elite horse is. There are a number of ways that we could have done this but in terms of what was reproducible and significant, for the purpose of our preliminary studies our elite horses were Grade One winners that earned a peak Beyer Speed figure of 108 or more. This group is in fact the elite of the elite in many ways and this population included the likes of Azeri, Commentator, Cigar, Caller One, Formal Gold, Serena’s Song and Turkoman. It included 10 of the past 18 winners of the Eclipse Award for Champion Racehorse and many household names so in terms of elite, the initial group of 200 horses was certainly that. The non-elite horses were also selected. These horses, in the main, were well bred horses by the likes of A.P Indy, Distorted Humor, Storm Cat, etc but they were slow. They had to have had at least 5 starts, no known impediments to their level of performance (surgeries, etc) and have not run a peak 78 Beyer Speed figure or better. After defining the model, we had to both train and then test the model. The training of the Performance Genetics model was completed on some 400 horses, 200 elite racehorses and 200 non-elite.