Estimating variance components from fullsib group means

Estimating heritability using family-pooled phenotypic and genotypic data: a simulation study applied to aquaculture

Estimating heritability based on individual phenotypic and genotypic measurements can be expensive and labour-intensive in commercial aquaculture breeding. Here, the feasibility of estimating heritability using within-family means of phenotypes and allelic frequencies was investigated. Different numbers of full-sib families and family sizes across ten generations with phenotypic and genotypic information on 10 K SNPs were analysed in ten replicates. Three scenarios, representing differing numbers of pools per family (one, two and five) were considered. The results showed that using one pool per family did not reliably estimate the heritability of family means. Using simulation parameters appropriate for aquaculture, at least 200 families of 60 progeny per family divided equally in two pools per family was required to estimate the heritability of family means effectively. Although application of five pools generated more within- and between- family relationships, it reduced the number of individuals per pool and increased within-family residual variation, hence, decreased the heritability of family means. Moreover, increasing the size of pools resulted in increasing the heritability of family means towards one. In addition, heritability of family mean estimates were higher than family heritabilities obtained from Falconer’s formula due to lower intraclass correlation estimate compared to the coefficient of relationship.

Prediction of breeding values for group-recorded traits including genomic information and an individually recorded correlated trait

Records on groups of individuals could be valuable for predicting breeding values when a trait is difficult or costly to measure on single individuals, such as feed intake and egg production. Adding genomic information has shown improvement in the accuracy of genetic evaluation of quantitative traits with individual records. Here, we investigated the value of genomic information for traits with group records. Besides, we investigated the improvement in accuracy of genetic evaluation for group-recorded traits when including information on a correlated trait with individual records. The study was based on a simulated pig population, including three scenarios of group structure and size. The results showed that both the genomic information and a correlated trait increased the accuracy of estimated breeding values (EBVs) for traits with group records. The accuracies of EBV obtained from group records with a size 24 were much lower than those with a size 12. Random assignment of animals to pens led to lower accuracy due to the weaker relationship between individuals within each group. It suggests that group records are valuable for genetic evaluation of a trait that is difficult to record on individuals, and the accuracy of genetic evaluation can be considerably increased using genomic information. Moreover, the genetic evaluation for a trait with group records can be greatly improved using a bivariate model, including correlated traits that are recorded individually. For efficient use of group records in genetic evaluation, relatively small group size and close relationships between individuals within one group are recommended.

Estimation of heritability and breeding values for early egg production in laying hens from pooled data

Under commercial conditions, data on egg production in laying hens are usually collected per cage rather than individually. In current breeding programs, genetic evaluations are, however, based on individually recorded egg production. Because commercial flocks are not maintained in single cages, this environmental difference between the breeding and commercial setting may result in a genotype x environment interaction. This study was aimed at estimating genetic parameters and predicting estimated breeding values for early egg production of laying hens by using pooled data (i.e., data from multiple bird cages) from pedigree birds housed in 4-bird cages. Using cage records, we compared 2 different methods of handling pooled data: cage sums and the assignment of cage means to individual birds, referred to as the approximate method. The 2 methods were compared by using cross-validation. Data from 3 purebred White Leghorn layer lines were used. Estimated heritability for early egg production was 0.36 when cage sums were used and 0.30 with the approximate method. The correlation of estimated breeding values between the cage sums method and the approximate method was 0.88. Cross-validation showed that the use of cage sums led to better predictions of missing phenotypes compared with the approximate method. The results of the research demonstrate that pooled data can be used in the genetic evaluation of laying hens and show that using directly pooled records (e.g., cage sums) gives better results than assigning group means to the birds of the group, thus simulating individual records.

Estimation of variance components and prediction of breeding values using pooled data

Data on performance of animals are, in several situations, collected at the group rather than individual level. Genetic evaluations in farm animals, however, are based on phenotypic information collected at the individual level. Therefore, it would be very attractive to extend genetic evaluations by incorporating information collected at the group level. In this paper we show the use of data collected at the group level for the estimation of variance components and the prediction of breeding values. We outline a general procedure that can be applied in different farm animal species. In the present work this procedure was applied to BW, for which pooled, as well as individual, observations were available, thus allowing for a comparison of the estimates, and to egg production, for which only pooled data were available. For BW at 19 and 27 wk the estimated heritabilities based on individual observations were very similar to those based on pooled observations. For BW at 43 and 51 wk, heritability estimates based on individual and pooled data were different, which can be caused by the emergence of competition effects. The accuracy of EBV predicted from pooled observations was about 70 to 80% of the accuracy of EBV predicted from individual observations. This result quantifies the loss deriving from the use of pooled instead of individual observations. Results show that estimation of variance components and breeding values from pooled data instead of individual observations is theoretically and practically feasible.

Estimation of genetic parameters based on individual and group mean records in laying hens

(1) The study was conducted to estimate the heritability, genetic correlations and breeding values of laying hens based on individual records and group mean records. (2) Records of two pure lines from a commercial breeding programme of White Leghorns from three generations housed in single cages and in group cages were used. A total of 8483 and 8817 individual records of lines A and D, respectively, and a total of 1358 (line A) and 1161 (line D) group mean records were analysed. (3) An animal model using Restricted Maximum Likelihood (REML) was used to estimate variance components of individual records. Group mean records were analysed using the sire model, taking heterogeneity of error variance and correlated residual effects into account. Breeding values of sires were estimated based on the BLUP method using a multivariate sire model. Spearman Rank correlations were used to compare sire breeding values estimated from individual records and from group mean records. The traits studied were monthly egg production, cumulative production and egg weight. (4) Heritability estimates based on individual records were higher than from group mean records. Heritabilities for cumulative production records were higher than for monthly production, based on individual as well as group mean records. The estimates of genetic correlations between monthly egg production and cumulative production were moderate to high. Egg production and egg weight recorded individually were highly genetically correlated with those recorded on group means. Sire breeding values estimated from individual records showed high correlations with those from group mean records. (5) Differences in the ranking of sire breeding values estimated from individual vs group mean records were negligible, indicating that no genotype x environment interaction exists. Selection based on individual performance records of laying hens housed in single cages could give a good response on performance of laying hens housed in group cages. Cumulative egg production over periods 1 to 6 is the best trait for the selection programme.