Effect of different levels of pedigree information and heritability magnitude on the quality of estimates of genetic parameters

Abstract EN

Milk production of the first (Milk1) and second (Milk2) parities was simulated according to the genetic and environmental variance-covariance structure of the two traits.

Genetic and environmental variances and covariances between the two traits were 202956, 275398, 216091 and 468931, 692353, 183786, respectively.

The simulated heritability (h2) estimates for the two traits were 0.3 and 0.28 and simulated genetic correlation (gc) 0.91.

Four levels of missing-sire pedigree information (m) were considered (complete, 10%, 30% and 50% missing-sire pedigree).

Two types of analyses (t) were performed, for the first trait Milk1, single and multiple traits analysis.

In the case of single trait analysis, the real variance components of this trait were modified to simulate three levels of (h2) (0.1, 0.3 and 0.5).

Twenty samples were simulated for each of 12 h2-m combinations.

The effect of different levels of m on the estimates of h2 and gc was also considered.

Twenty samples were simulated in the case of multiple traits analysis (Milk1 and Milk2) for each level of m.

The simulated heritability’s and gc were 0.3, 0.28 and 0.91 (the real values), respectively.

MTDFREML program was used to estimate h2 and gc in different cases.

Mean squared error (MSE) and bias were used as criteria to evaluate the accuracy of different studied cases.

In the case of single trait analysis the smallest values of MSE (0.0221, 0.0441) were observed at h2=0.1 and at m=10%, respectively.

As h2 increased both bias and MSE significantly increased.

The effect of m levels was not significant on bias but significant (P < 0.05) on MSE estimates where it generally increased with m.

Considering the gc, the least estimate of MSE was obtained at the 10% level of m and the least accurate one was at the highest level of m.

The effect of m was not significant (p > 0.05) on the estimates of h2, bias and mean squared error estimates, whereas the effect of t was significant (p < 0.05) on both of h2 and bias estimates.

Results of single trait analysis for Milk1 showed that the trait with low heritability (0.1) and low amount of m yielded the most accurate estimate of h2, whereas the results of multiple traits analyses indicated that as the amount of m increases, the estimates of h2 and genetic correlation become far from the true values.

Type of analysis only affected significantly the estimates of h2 and bias estimates but did not affect the estimates of MSE under the circumstances of this study.