Bayesian estimation of genetic parameters and comparison of statistical models for race performance traits in the Turkmen horse.

Overview

• This study estimated the genetic influence on race performance traits in Turkmen horses using Bayesian methods and compared different statistical models to find the most suitable approach for genetic evaluation.
• The research highlights the importance of selecting appropriate models that consider various race-specific and management factors to improve the accuracy of genetic evaluations in racehorses.

Background and Importance

• Race performance in horses is a key factor for breeding decisions but is influenced by both genetic factors and a wide range of non-genetic influences.
• Non-genetic factors include race conditions, horse management, handicaps, and the skill or influence of the jockey, making performance a complex trait to analyze.
• Different ways of modeling these non-genetic influences can lead to different estimates of genetic parameters such as heritability and breeding values.
• Accurate genetic evaluation helps breeders select horses that will likely pass on desirable racing traits, improving the racing stock over time.

Data and Traits Analyzed

• The study used a large dataset of 10,830 official race records collected over 20 years (2003-2023).
• The dataset included performance from 1,667 Turkmen horses (899 stallions and 768 mares) bred in Iran.
• Three specific race performance traits were analyzed:
  • Race Completion Time (RCT)
  • Rank at the End of Competition (REC)
  • Average Speed (AS)

Statistical Models and Bayesian Estimation

• Eight different hierarchical models were tested, each with unique random-effect structures accounting for various sources of variation linked to horses, races, jockeys, and other factors.
• Bayesian estimation methods were used to estimate genetic parameters, such as heritability, which indicate the proportion of observed variation in traits that can be attributed to additive genetic effects.
• Model 8 (M8) was identified as the best-fitting model for all three traits based on statistical support.

Key Results

• Heritability estimates were low for all traits:
  • RCT: 0.006 ± 0.001
  • REC: 0.06 ± 0.01
  • AS: 0.004 ± 0.001
• Prediction bias ranged from 0.12 to 0.20, indicating some deviation between predicted and observed values but not severe.
• Prediction accuracy was relatively moderate to high, between 0.60 and 0.78, showing the models could reliably predict performance to a degree.
• The low heritability values suggest that most of the variation in performance traits is due to non-genetic factors and environment rather than additive genetics in this population.

Interpretation and Implications

• The study emphasizes that genetic evaluation models must include important non-genetic effects such as race conditions and jockey influence because these contribute substantially to performance outcomes.
• Incorporating these effects improves estimation of genetic parameters and breeding values, reducing bias caused by ignoring such complex environmental influences.
• Although this research focused on Turkmen horses in Iran, the methodological insights can be applicable to other racing horse populations experiencing complex and multifactorial influences on performance.
• Breeding programs can utilize such tailored models to more accurately identify genetically superior racehorses, even when heritability is low.

Conclusion

• This research demonstrates the value of Bayesian methods and sophisticated statistical modeling in dissecting genetic and environmental components of race performance.
• Appropriate model specification that accounts for race- and jockey-specific effects is crucial for reliable genetic evaluation in racehorses.
• The findings highlight that, in populations where environmental influences dominate, careful modeling is necessary to avoid overestimating genetic contributions to racing performance.