Multiple trait and random regression models using linear splines for genetic evaluation of multiple breed populations

Area of Science:

• Animal Genetics
• Quantitative Genetics
• Statistical Genetics

Background:

• Accurate genetic parameter estimation is crucial for effective animal breeding programs, especially in diverse, multiple-breed populations.
• Traditional models may not fully capture complex genetic structures and interactions within admixed populations.

Purpose of the Study:

• To evaluate the suitability of random regression models using linear splines (RRMLS) for estimating genetic parameters in multiple-breed populations.
• To investigate potential interactions between sire breeding values (BV) and their progeny's breed group composition.

Main Methods:

• Simulated ten populations by crossing two breeds, analyzing genetic parameters using a multiple-trait model (MULT) and RRMLS.
• Applied RRMLS with varying knot numbers (3, 5, 7) to field data for Holstein-Gyr cattle, analyzing age at first calving (AFC), lactation length (LL), and 305-day milk yield (MY-305).

Main Results:

• RRMLS and MULT models provided similar, accurate genetic parameter estimates in simulations, with high correlations (0.74–0.76) between simulated and estimated BVs.
• RRMLS with 7 knots best fit AFC and LL, while 5 knots best fit MY-305, indicating optimal model complexity varies by trait.
• Heritability estimates ranged from 0.10 to 0.48 across traits, and a significant interaction between sire BV and progeny breed group was detected.

Conclusions:

• RRMLS is a robust and useful tool for genetic evaluation in complex, multiple-breed populations.
• Genetic parameter estimates for traits like AFC, LL, and MY-305 are influenced by the breed composition of the progeny.
• The identified interaction necessitates considering breed composition in breeding value predictions for admixed populations.