Genetic parameters and principal component analysis for egg production from White Leghorn hens

Genetic parameters and principal component analysis for egg production from White Leghorn hens

Genetic parameters and principal component analysis for egg production from White Leghorn hens G. C. Venturini,* R. P. Savegnago,* B. N. Nunes,* M. C...

821KB Sizes 0 Downloads 13 Views

Genetic parameters and principal component analysis for egg production from White Leghorn hens G. C. Venturini,* R. P. Savegnago,* B. N. Nunes,* M. C. Ledur,† G. S. Schmidt,† L. El Faro,‡ and D. P. Munari*1 *Departamento de Ciências Exatas, Faculdade de Ciencias Agrarias e Veterinarias (FCAV)/ Universidade Estadual Paulista (UNESP), Jaboticabal, São Paulo, Brazil 14884-900; †Embrapa Suínos e Aves, Concórdia, Santa Catarina, Brazil 89700-000; and ‡Agência Paulista de Tecnologia dos Agronegócios (APTA) Centro Leste/Secretaria de Agricultura e Abastecimento (SAA), Ribeirão Preto, São Paulo, Brazil 14001-970 ABSTRACT The objectives of this study were to estimate genetic parameters for accumulated egg production over 3-wk periods and for total egg production over 54 wk of egg-laying, and using principal component analysis (PCA), to explore the relationships among the breeding values of these traits to identify the possible genetic relationships present among them and hence to observe which of them could be used as selection criteria for improving egg production. Egg production was measured among 1,512 females of a line of White Leghorn laying hens. The traits analyzed were the number of eggs produced over partial periods of 3 wk, thus totaling 18 partial periods (P1 to P18), and the total number of eggs produced over the period between the 17 and 70 wk of age (PTOT), thus totaling 54 wk of egg production. Estimates of genetic parameters were ob-

tained by means of the restricted maximum likelihood method, using 2-trait animal models. The PCA was done using the breeding values of partial and total egg production. The heritability estimates ranged from 0.05 ± 0.03 (P1 and P8) to 0.27 ± 0.06 (P4) in the 2-trait analysis. The genetic correlations between PTOT and partial periods ranged from 0.19 ± 0.31 (P1) to 1.00 ± 0.05 (P10, P11, and P12). Despite the high genetic correlation, selection of birds based on P10, P11, and P12 did not result in an increase in PTOT because of the low heritability estimates for these periods (0.06 ± 0.03, 0.12 ± 0.04, and 0.10 ± 0.04, respectively). The PCA showed that egg production can be divided genetically into 4 periods, and that P1 and P2 are independent and have little genetic association with the other periods.

Key words: breeding value, genetic correlation, heritability, multivariate analysis 2013 Poultry Science 92:2283–2289 http://dx.doi.org/10.3382/ps.2013-03123

INTRODUCTION One of the main objectives of selection in genetic breeding programs for laying hens is egg production. This trait can be measured as an absolute number or as the percentage of eggs produced over a given period. According to Ledur et al. (1993), estimates of genetic and phenotypic parameters are the main tools for choosing the selection method and the traits that should be considered as selection criteria for obtaining genetic gains in egg production. Selection of egg-laying birds based on partial egg production can increase total egg production (Poggenpoel et al., 1996). According to Silva et al. (1984), the interval of generations per time unit can be reduced by half if selection is done taking partial egg production into ©2013 Poultry Science Association Inc. Received February 18, 2013. Accepted May 13, 2013. 1 Corresponding author: [email protected]

consideration. However, it is important to estimate genetic parameters both for partial periods of egg production and for total egg production because estimates of genetic parameters are inherent to a given population and may vary due to differences in the genetic makeup and management of the population in question, compared with other results in the literature. Principal component analysis (PCA) is one of the most used multivariate techniques. It is a mathematical procedure that uses orthogonal transformations of the covariance matrix with the aim of reducing a set of correlated variables to a smaller number of noncorrelated variables called principal components, with minimal loss of original information from the covariance matrix (Hair et al., 2009). When this technique is applied using the breeding values of the traits evaluated, it is possible to explore the most important genetic relationships in the most relevant principal components (i.e., those that explain the majority of the information in the database) and also to demonstrate relationships that are not observable directly through

2283

2284

Venturini et al.

the genetic correlation matrix. Savegnago et al. (2011) applied PCA to breeding values for traits related to egg production in White Leghorn hens and found different relationships among the breeding values of the traits depending on the dimension evaluated (i.e., in relation to the principal component analyzed). When the breeding values of several traits are used in PCA, they can be considered to be genetic indices that summarize all the traits evaluated, in addition to being an approach for exploring the genetic relationships among the traits. Buzanskas et al. (2013) used this approach to build selection indices, in which the genetic weighting factors were obtained through PCA using the breeding values of reproductive traits and BW among female Canchim beef cattle. The objectives of this study were to estimate genetic parameters for accumulated egg production over 3-wk periods and for total egg production over 54 wk of egg laying, and using PCA, to explore the relationships among the breeding values of these traits to identify the possible genetic relationships present among them and hence to observe which of them could be used as selection criteria for improving egg production.

sidering the total period) were also excluded to eliminate the possibility of including birds with subclinical diseases that could affect egg production (Fairfull and Gowe, 1990). Estimates of the genetic parameters were obtained through the restricted maximum likelihood method, using 2-trait animal models, by means of the MTDFREML (multiple trait derivative-free restricted maximum likelihood) software (Boldman et al., 1995). The initial values used to carry out the iteration process were based on results found in the literature (Anang et al., 2000). The fixed effect of hatch was significant (P < 0.01) for egg production over both the partial and the total periods and age of first egg was significant (P < 0.01) for P1, P2 and P3, using the GLM procedure from SAS (SAS 9.2, SAS Institute Inc., Cary, NC). The general model included the fixed effect of hatch, the covariate age of first egg (only for P1, P2, and P3) and the random additive genetic and residual effects. In matrix form,

MATERIALS AND METHODS

where X and Z are, respectively, n × p and n × q incidence matrices; and β, u, and e are the fixed effect, genetic, and residual random effects vectors, respectively. The classification of the breeding values predicted for the 43 sires for P3 to P12 was compared with the classification for PTOT, using Spearman’s rank correlation. The objective of this analysis was to observe whether the selection based on partial records of up to 53 wk of age would be efficient for improving total production. The PCA was carried out using the breeding values for the 3-wk periods (EBV-P1 to EBV-P18) and the breeding values for the total egg production (EBVPTOT). This analysis had the objective of condensing the original information contained in these 19 variables into a smaller number of latent and orthogonal variables named principal components, with minimum loss of information (Hair et al., 2009). In addition, the analysis had the objective of exploring the relationships among the breeding values of these traits to observe possible relationships that were not directly observable through the genetic correlation matrix. Breeding values for PTOT (EBV-PTOT) were used as a supplemental variable; that is, this variable had no influence on the correlation matrix used for construction of the principal components. It was not used as an active variable because EBV-PTOT presented information already included in other variables. The reason for adding a supplemental variable in this analysis was to help the interpretation of results. Thus, it is possible to interpret the relationship between the breeding values of the partial periods in each principal component with EBV-PTOT. To carry out the PCA, the breeding values of the traits assessed were standardized using the normal standard distribution (z scores).

The data set used in the present study originated from a population of White Leghorn hens, named CC, that had been developed and maintained by the Embrapa (Brazilian Agricultural Research Corporation) Swine and Poultry National Research Center, Concórdia, Santa Catarina, Brazil. The CC birds were from a pure line selected mainly for their egg production, egg weight, food conversion, hatchability, sexual maturity, fertility, egg quality, and reduced BW (Figueiredo et al., 2003; Rosário et al., 2009). Data on 1,512 birds that originated from 3 hatches were analyzed at intervals of 15 d. The egg production of each individual was measured from the numbers of eggs gathered between 17 and 70 wk of age. In each week, eggs were gathered on 5 d of the week. According to Wheat and Lush (1961), this measurement has a 0.99 correlation with the egg production on 7 d of the week. The traits analyzed were the number of egg produced over partial periods of 3 wk, thus totaling 18 partial periods (P1 to P18), and the total number of eggs produced over the period between the 17 and 70 wk of age (PTOT), thus totaling 54 wk of egg production. The pedigree file contained 2,128 birds. The structure of the pedigree data of the generation studied consisted of 43 sires mated with 232 hens in a hierarchical scheme (at least 5 females per male) by means of artificial insemination. The sires had at least 12 daughters. The dams with fewer than 2 daughters and those ones that died during the production cycle were excluded from the data file. The females that had total production of fewer than 63 eggs (30% of the egg-laying rate, con-



y = Xβ + Zu + e,

2285

EGG PRODUCTION FROM WHITE LEGHORN HENS

The choice of principal components to explain the majority of the additive genetic variation of the data set was determined as those with eigenvalues greater than one, in accordance with Kaiser’s criteria (Kaiser, 1960). The eigenvalue of a principal component is associated with the variance of all the traits included in the principal component. Each eigenvalue is associated with a unit vector named an eigenvector (Rencher, 2002). The eigenvectors represent the magnitude, direction, and sense of the vector in relation to the importance of each trait within the principal component. The first principal component explains the greatest part of the total additive genetic variance. The second principal component explains the second greatest part of the total additive genetic variance and so on, until all of the variance is explained. In a data set with p variables, the ith principal component (PCi) is given by PCi = ai1X1 + ai2X2 + · · · + aijXj, where i = 1, 2, …, 19 and j = 1, 2, …, 19, aij is the jth standardized coefficient of the jth variable in the ith principal component, and Xj is the jth value of the original variable. The standardized coefficients are calculated by

aij =

eigenvectorij eigenvalue j

,

where aij is the standardized coefficient for the breeding values of the ith variable in the jth principal component. The PCA was processed using the Statistica 8.0 software (Statistica 8.0, Statsoft Inc., Tulsa, OK).

The heritability estimates for partial egg production in the present study were smaller than those found by Wei and Van der Werf (1993), which were 0.41 to 0.49 (between 18 and 25 wk of age), 0.21 (between 26 and 65 wk of age), and 0.40 (between 18 and 65 wk of age), probably due to the different methodologies used to estimate the genetic parameters. Sabri et al. (1999) reported heritability estimates of 0.27 ± 0.17, 0.19 ± 0.19, and 0.30 ± 0.07 for partial egg production from 26 to 30, 50 to 54, and 26 to 54 wk of age from White Leghorn birds, which correspond respectively to the P4 to P5, P12 to P13, and P4 to P13 periods of the present study. Anang et al. (2000) found heritability estimates for partial and total egg production of 0.18 ± 0.02 to 0.49 ± 0.02. The genetic, environmental, and phenotypic correlations estimated between PTOT and the egg production of the 3-wk partial periods (P1 to P18), obtained through 2-trait analyses, are shown in Figure 2. The greatest genetic correlations were between PTOT and the partial periods P10, P11, and P12, whereas the smallest was between PTOT and P1. Anang et al. (2000) reported genetic correlations between the monthly egg productions and the total accumulated production ranging from 0.06 to 0.73, which were close to those found in the present study. The genetic correlations between P2 to P18 and PTOT ranged from 0.64 ± 0.18 (P2) to 1.00 ± 0.10 (P10 and P12; Figure 2). The low genetic correlation between P1 and PTOT (0.19 ± 0.31) indicated that this partial production brings no gain for total egg production within the laying cycle. In addition, selection of total production based on the egg production of the first weeks may cause a decrease in the age of sexual maturity, which would cause an increase in the frequen-

RESULTS AND DISCUSSION The means for partial egg production were practically constant throughout the production period (Table 1), except in P1 (from 17 to 19 wk of age), which was the lowest due to the great variation in the age first egg, (i.e., the age of the birds, measured in days, when the first egg laying occurred). Age at first egg was at 20 wk of age, which corresponded to the period P2. The mean and respective SD for the egg production during this period were 10.12 ± 4.12. The heritability estimates for egg production over the 3-wk partial periods and the total period, obtained through 2-trait analyses (Figure 1), ranged from 0.05 ± 0.03 (P1 and P8) to 0.27 ± 0.06 (P4). The heritability estimates for PTOT, obtained through 2-trait analysis on each partial period, did not present great differences among each other, ranging from 0.22 ± 0.06 (P8, P9, P10, P12, and P13) to 0.25 ± 0.06 (P15 and P16), and were similar to those estimates found by Munari et al. (1992) and Ledur et al. (1993), who used the least squares method to estimate genetic parameters.

Table 1. Means and respective SD and CV for the 3-wk (partial periods P1–P18) and total (PTOT) egg production Egg production period

Age (wk)

Mean

SD

CV (%)

P1 P2 P3 P4 P5 P6 P7 P8 P9 P10 P11 P12 P13 P14 P15 P16 P17 P18 PTOT

17 20 23 26 29 32 35 38 41 44 47 50 53 56 59 62 65 68 17

1.91 10.12 13.10 12.99 13.41 13.30 13.11 13.25 13.04 12.85 12.62 12.38 12.16 11.79 11.58 11.12 10.66 10.15 209.77

2.84 4.12 2.49 2.46 2.43 2.72 2.98 2.87 2.80 2.96 3.11 3.22 3.28 3.45 3.59 3.81 3.87 3.94 29.81

149.14 40.71 19.02 18.98 18.14 20.49 22.70 21.65 21.50 23.00 24.67 26.02 27.01 29.28 31.05 34.26 36.30 38.82 14.21

to to to to to to to to to to to to to to to to to to to

19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 70

2286

Venturini et al.

Figure 1. Heritability estimates for the accumulated egg production in the 3-wk periods (partial periods P1 to P18) and the total egg production over the 54 wk of laying (PTOT), obtained through the 2-trait animal model.

cy of small eggs due to the incomplete sexual development of the very young hens (Savegnago et al., 2011). Spearman’s correlation on the breeding value classifications of the sires between PTOT and P10, P11, and P12 was significant (P < 0.01). It was 0.99 between PTOT and P11 and 1.00 between PTOT and P10 and P12, indicating that the ranking of each bird was practically the same when classified through the breeding values of these partial periods or through the total period. However, despite the significance of the estimates of Spearman’s correlation and the perfect genetic correlation (equal to 1) between PTOT and the partial periods of P10 and P12, the heritability estimates for P10, P11, and P12 were very low in relation to PTOT. Moreover, P10, P11, and P12 are close to the end of the egg production cycle. So, the selection criterion using these partial periods could disregard the genes expres-

sion related to natural molting and broodiness, which are related to late partial egg production periods and could cause a decrease in the performance of total egg production when the entire egg cycle is not considered in breeding programs (Savegnago et al., 2011). Then, it is better to use the total egg production instead of late partial egg production even though the genetic correlations were 1. Thus, bird selection based on P10, P11, and P12 would not be efficient for improving PTOT. According to Schmidt et al. (1996) and Savegnago et al. (2011), the effects of genes relating to persistency of egg laying, natural molting and brooding, expressed from the middle to the end of the egg production cycle, are not taken into account in the selection process for total egg production, if the bird selection is based on partial periods at the beginning of the laying cycle (approximately

Figure 2. Estimates of genetic, residual, and phenotypic correlations between the accumulated egg productions in the 3-wk periods (partial periods P1 to P18) and the total egg production over the 54 wk of laying (PTOT), obtained through the 2-trait animal model.

EGG PRODUCTION FROM WHITE LEGHORN HENS

until 40 wk), which may cause a decrease in the total egg production performance of the birds. Genetic parameters of traits that are measured over the course of time, as is the case of egg production, can be estimated by means of random regression models (Anang et al., 2002; Luo et al., 2007; Wolc and Szwaczkowski, 2009; Wolc et al., 2011; Venturini et al., 2012). These models have the advantage of taking into account the covariance structure for genetic and environmental changes for a trait over time. This makes the genetic parameter estimates more precise, thereby avoiding over- or underestimates. Venturini et al. (2012) used random regression models on the same data set to estimate the genetic parameters of egg production over a 3-wk period and observed that the heritability estimates of the first 12 wk (P1 to P4) were smaller than those reported in the present study. For the other periods, the heritability estimates were close, using both the random regression models and the 2-trait animal model. In accordance with Kaiser’s criteria, 3 principal components were chosen: PC1, PC2, and PC3 (Figure 3). From the 19 original dimensions (EBV-P1 to EBV-P18 and EBV-PTOT), 70.93% of the variance of the breeding values was explained in these 3 principal components, of which 45.19% was in PC1, 16.41% in PC2, and 9.33% in PC3. In PC1, the vector projection of all the 3-wk breeding values, except EBV-P1 and EBV-P2, presented an association with EBV-PTOT, indicating that the highest breeding values for total egg production are associated with the highest breeding values for the partial periods (Figure 4). The PC2 was a dimension that did not greatly discriminate among the breeding values for total egg production because the EBV-PTOT projection in PC2 was close to the meeting of the axes (close to 0). In PC3, EBV-P1 and EBV-P2 had a very low genetic association with the other partial periods and

2287

with EBV-PTOT, and this was also observed through the small size of the EBV-P1 and EBV-P2 vectors in the PC1 versus PC2 graph. According to Savegnago et al. (2011), when the genetic correlation between 2 traits tends to 0, the breeding values of the traits are separated in different principal components, as observed with EBV-P1 and EBV-P2, which had greater expression in PC3 because of the low genetic correlation of egg production between these periods and PTOT (Figure 2). There were 4 groups of variables in the PC1 vs. PC2 graph (Figure 4a). The first group encompassed EBVP1 and EBV-P2; the second, EBV-P3 to EBV-P9; the third, EBV-P10 to EBV-P15 and EBV-PTOT; and the fourth, EBV-P16 to EBV-P18. This indicated that, genetically, egg production could be divided into 4 stages. So, bird selection based on 3-wk laying periods is not recommended for improving total egg production, based on the magnitude of the heritability estimates. Bird selection must be based on total egg production. Principal component analysis was useful for observing how the laying cycle can be divided and what the relationships of the breeding values of egg production are between the partial periods and the total period.

ACKNOWLEDGMENTS Financial support was provided by the Brazilian Agricultural Research Corporation (Embrapa; Empresa Brasileira de Pesquisa Agropecuária). G. C. Venturini and B. N. Nunes received scholarships from the Coordination Office for Advancement of University-level Personnel (Coordenação de Aperfeiçoamento de Pessoal de Nível Superior) in conjunction with the Postgraduate Program on Genetics and Animal Breeding, Faculdade de Ciências Agrárias e Veterinárias, Universidade Estadual Paulista. R. P. Savegnago was granted scholarships by the São Paulo Research Foundation (Fundação

Figure 3. Eigenvalues (y-axis) and percentage of the original variation stored in each of the 19 principal components (x-axis).

2288

Venturini et al.

Figure 4. Principal component analysis with the breeding values of the partial (EBV-P1 to EBV-P18) and total egg production (EBV-PTOT). a. Principal component 1 vs. 2 (PC1 vs. PC2). b. Principal component 1 vs. 3 (PC1 vs. PC3).

EGG PRODUCTION FROM WHITE LEGHORN HENS

de Amparo à Pesquisa do Estado de São Paulo). L. El Faro and D. P. Munari held productivity research fellowship from the National Council for Scientific and Technological Development (Conselho Nacional de Desenvolvimento Científico e Tecnológico).

REFERENCES Anang, A., N. Mielenz, and L. Schüler. 2000. Genetic and phenotypic parameters for monthly egg production in White Leghorn hens. J. Anim. Breed. Genet. 117:407–415. Anang, A., N. Mielenz, and L. Schüler. 2002. Monthly model for genetic evaluation of laying hens. II. Random regression. Br. Poult. Sci. 43:384–390. Boldman, K. G., L. A. Kriese, L. D. Van Vleck, C. P. Van Tassel, and S. D. Kachman. 1995. A Manual for the use of MTDFREML. A Set of Programs to Obtain Estimates of Variance and Covariances (Draft). USDA-ARS, Lincoln, NE. Buzanskas, M. E., R. P. Savegnago, D. A. Grossi, G. C. Venturini, S. A. Queiroz, L. O. C. Silva, R. A. A. Torres Júnior, D. P. Munari, and M. M. Alencar. 2013. Genetic parameter estimates and principal component analysis of breeding values of reproduction and growth traits in female Canchim cattle. Reprod. Fertil. Dev. 25:775–781. Fairfull, R. W., and R. S. Gowe. 1990. Genetics of egg production in chickens. Pages 705–759 in Poultry Breeding and Genetics. R. D. Crawford, ed. Elsevier Science Publishers, Amsterdam, the Netherlands. Figueiredo, E. A. P., G. S. Schmidt, M. C. Ledur, V. S. Avila, P. A. R. Brum, L. Fiorentin, and F. R. F. Jaenisch. 2003. Genetic gain in egg production and egg weight in White Leghorn Embrapa 011. Proc 9th World Conf. Anim. Prod., Porto Alegre, Brazil. Hair, J. F., W. C. Black, B. J. Babin, and R. E. Anderson. 2009. Multivariate Data Analysis. 7th ed. Prentice Hall, Upper Saddle River, NJ. Kaiser, H. F. 1960. The application of electronic computers to factor analysis. Educ. Psychol. Meas. 20:141–151. Ledur, M. C., G. S. Schmidt, E. A. P. Figueiredo, V. S. Avila, and L. Balen. 1993. Genetic and phenotypic parameters for productive traits in white egg layer stocks. Pesquisa Agropecu. Bras. 28:1031–1037. Luo, P. T., R. Q. Yang, and N. Yang. 2007. Estimation of genetic parameters for cumulative egg numbers in a broiler dam line by using a random regression model. Poult. Sci. 86:30–36.

2289

Munari, D. P., G. S. Schmidt, and M. L. Giannoni. 1992. Variabilidade genética e fenotípica do desempenho de uma linhagem de aves Leghorn. I. Herdabilidades. Ars Vet. 8:49–57. Poggenpoel, D. G., G. F. Ferreira, J. P. Hayes, and J. J. Du Preez. 1996. Response to long-term selection for egg production in laying hens. Br. Poult. Sci. 37:743–756. Rencher, A. C. 2002. Methods of Multivariate Analysis. Wiley-Interscience, New York, NY. Rosário, M. F., M. C. Ledur, A. S. A. M. T. Moura, L. L. Coutinho, and A. A. F. Garcia. 2009. Genotypic characterization of microsatellite markers in broiler and layer selected chicken lines and their reciprocal F1s. Sci. Agric. 66:150–158. Sabri, H. M., H. R. Wilson, R. H. Harms, and C. J. Wilcox. 1999. Genetic parameters for egg and related characteristics of white leghorn hens in a subtropical environment. Genet. Mol. Biol. 22:183–186. Savegnago, R. P., S. L. Caetano, S. B. Ramos, G. B. Nascimento, G. S. Schmidt, M. C. Ledur, and D. P. Munari. 2011. Estimates of genetic parameters, and cluster and principal components analyses of breeding values related to egg production traits in a White Leghorn population. Poult. Sci. 90:2174–2188. Schmidt, G. S., E. A. P. Figueiredo, D. P. Munari, O. S. Rosa, and M. C. Ledur. 1996. Frequência de muda natural e choco em aves selecionadas para a produção de ovos. Conferência Apinco de Ciências e Tecnologia Avícolas. FACTA, Curitiba, Brazil. Silva, M. A., R. F. Euclydes, P. R. Soares, and J. B. Fonseca. 1984. Análise genética de características de importância econômica em poedeiras leves. R. Bras. Zootec. 13:82–94. Venturini, G. C., D. A. Grossi, S. B. Ramos, V. A. R. Cruz, C. G. Souza, M. C. Ledur, L. El Faro, G. S. Schmidt, and D. P. Munari. 2012. Estimation of genetic parameters for partial egg production periods by means of random regression models. Genet. Mol. Res. 11:1819–1829. Wei, M., and J. H. Van der Werf. 1993. Animal model estimation of additive and dominance variances in egg production traits of poultry. J. Anim. Sci. 71:57–65. Wheat, J. D., and J. L. Lush. 1961. Accuracy of partial trapnest records. 1. Repeatability of daily egg records. Poult. Sci. 40:399– 406. Wolc, A., J. Arango, P. Settar, N. P. O’Sullivan, and J. C. M. Dekkers. 2011. Evaluation of egg production in layers using random regression models. Poult. Sci. 90:30–34. Wolc, A., and T. Szwaczkowski. 2009. Estimation of genetic parameters for monthly egg production in laying hens based on random regression models. J. Appl. Genet. 50:41–46.