Predicting breeding values in Murrah buffaloes using random regression models

Abstract

Various random regression models with different order of Legendre polynomials for permanent environmental and genetic effects were constructed to evaluate Murrah buffaloes under field condition. A total of 24,160 test-day (TD) milk yield records of 2,336 buffaloes belonging to six districts of Haryana state from 2014 to 2019 under CHRS Rohtak were considered for estimating the random covariate coefficients using REML algorithm. The R2 and BIC improved up to seventh order of the polynomials but decreased at 8th order. The lowest value of AIC was for 8th order of polynomial. The models with LPs of order 5 and above gave an improved fit of the lactation curve having high R2 value (>98%). However, the number of parameters increased from 13 for LP of order 3 to 73 for LP of order 8 indicating over parameterization of models. Best order of fit was 7th order having highest R2 (99.23%), lowest AIC and lowest BIC value. The lactation curves plotted from the ETDMYs obtained under RRMs using LP5 and LP7 were smoother as compared to the lactation curve obtained from arithmetic mean. The 301 day lactation yield under RRMs using LP of order 7 and 5 averaged 3470.42±2.76 kg and 3460.72±3.10 kg, respectively. The milk yield ranged from 3,456.38 kg in Bhiwani district to 3,480.86 kg in Sonipat district under LP7 model while it ranged from 3,426.35 kg in Jhajjar district to 3,503.95 kg in Bhiwani district under LP5 model. It was observed that less than 50% were common in top 1% of total buffaloes under the two RRMs. The RRM using LP of order 7, which provided more accurate breeding values, must be considered for selection of top ranking/ superior animals for breeding purpose.

Keywords

• Breeding value
• Murrah buffaloes
• Prediction
• Random regression model