Fine-tuning of season definition for genetic analysis of fertility, productivity, and longevity traits in Iranian Holstein dairy cows

Abstract

Fixed environmental effects have shown to affect random genetic and residual effects.In this study, we evaluated various month merge classes as fixed environmental effecton estimation of genetic parameters for production, reproduction and longevity traitsof Iranian Holstein dairy cows. Data were collected from Holstein cows, in Isfahanprovince of Iran from 1993 to 2009. First, the edited data (53,908 records) were analyzedusing general linear model (GLM) in SAS package. Single-month classes yielded theleast mean square error (MSE) and the highest R-square. Then, the restricted maximumlikelihood (REML) and the best linear unbiased prediction (BLUP) procedures wereused to estimate genetic parameters and breeding values (EBVs) from the best models,which included single-month effect compared to triple-month (standard astronomicalseason in northern hemisphere) effect as a traditional model. Agreeing with the analysisof variance (ANOVA) results, standard season effect also led to higher MSE, Akaikeinformation criterion (AIC), Bayesian information criterion (BIC), and likelihood ratiotest (LRT). However, estimated heritabilities and subsequently mean accuracies forEBVs were considerably higher, when alternative definitions of season were explored.In conclusion, results of this study showed a considerable tradeoff between “best”(MSE, AIC, BIC, and LRT) and “unbiased” model indicators (estimated heritability andmean accuracy of EBVs). Importantly, this confounding effect was more evident forreproduction traits, age at first service (AFS) in particular. However, further worldwideevaluations of Holstein dairy cows are needed to determine the importance of modeloptimization on random effect predictions.