The implementation of principal component analysis in genetic evaluation of Iranian dairy cattle

In the quantitative genetics area, random regression model is one of the most accurate models for estimating daily breeding value in dairy cattle. However, because of the higher number records per each cow, application of this model is labor and time consuming. In addition, breeding values of cows at different days of lactation are highly correlated. The main objectives of the current study were to determine the relative importance of each breeding value at different days of lactation and to estimate the genetic principal components for the breeding values of Iranian Holstein dairy cattle for milk production traits. Maternal and methods: records of milk production traits of first-parity dairy cows. Milk yield, fat percentage and protein percentage test-day records of 73839, 65165 and 46881 cows, respectively, from 230 herds with 176390 cows in their pedigree were used in the analyses. Only test-day records belonging to 5 to 305 days of lactation were used. The data belonged to cows were born between 1988 and 2015 with age at first calving ranged between 21 to 48 m. In addition, the existence of at least one monthly record in the first 90 days after calving was essential for the cow, otherwise it would be eliminated. These data were collected by National breeding center, Karaj, Iran. Genetic parameters were estimated by a random regression model and Bayesian approach using GIBSS3F90 software. The estimated breeding values at all days of lactation were calculated and standardized using the standard score (z). Then, Correlation matrices among breeding values at different days of lactation and genetic principal components of breeding values were estimated by PROC CORR and PROC PRINCOMP of SAS software, respectively. Finally, we could calculate principal component score as a selection criterion (selection index) for the selection of dairy cattle. For this purpose, the standardized score coefficient was obtained by dividing the daily eigenvector of each principal component by square root of its eigenvalue. The principal component score were calculated of the sum of the multiply between standardized score coefficient and daily standardized breeding values for each cow. However, the principal components could be used as an index to multiple traits evaluation of animals. The genetic correlations matrix between the estimated breeding values at different days of lactation demonstrated that the breeding values at the middle stage of lactation were highly correlated with the breeding values at the reaming stages of lactation. The genetic principal component analysis revealed that the first two principal components accounted for a high percent of total genetic variance of all studied traits. For milk yield, the first principal component explained 99.48% of genetic variance, while two first components explained almost 98.19% and 100% of genetic variance for fat percent and protein percent traits, respectively. The absolute value of correlations between the first principal component of milk yield and all breeding values (except for day 56 and day 231) were more than 0.056. The absolute values of correlations between the first principal component of fat percent and the daily breeding values were greater than 0.06 for days between 83 and 222; and for protein percent were greater than 0.07 for days 99 to 168 and days 289 to 305.
Considering the high correlation between breeding values seem to, were estimated breeding values for all days is not required. The first principal component milk yield trait with nearly all estimated breeding values, high correlation and first two principal component fat percent trait of estimated breeding values in the early and middle of lactation period had a high relationship. But first two principal component protein percent trait of estimated breeding values in the middle and later of lactation period had a high correlation. Considering the high cost of recording system in dairy cattle industry and the high correlation between the breeding values, it seems that there is no need to predict the breeding value for all days of lactation. In other words, reducing the number of records per each cow may be beneficial at both economic and genetics stand points. Furthermore, due to the high, direct correlation between the principal components and daily breeding values, the implementation of principal components in the genetic merit evaluation of selection candidates for production-related traits is suggested.